The prompts behind the commits of this repository, as given to Claude Code, newest first, each with its commit and pull request. Square brackets mark what was left out or redacted, and a commit without a prompts section is left out. This file is generated from the commit messages, so a change by hand is lost at the next commit.
2026-10-28, the commit that last changed this file, model claude-opus-5-5, effort xhigh.
21.
dont recommend me a slack draft if i dont ask so. [an ask about a guide left out]
2026-10-28, commit 68ebf61, model claude-opus-5-5, effort xhigh.
5.
dont show things that havent changed just blank or delete whole row if even frontned didnt change. also cant you render tables in your responses?
(Claude asked how tables should show from now on)
Render to read
2026-10-28, commit fabc504, model claude-opus-5-5, effort xhigh.
4.
[two questions about a downstream project left out] also no need for the prompt since theres nothing [a teammate] should fix in their future work
(Claude asked whether skipping the teammate prompt holds from now on)
only when i say why does a problem x still exist in like the closing of issues was
2026-10-28, commit c3b6af0, model claude-opus-5-5, effort xhigh.
1. Quoted under 27b102b.
2. (answers the next-step question)
[a reply about a downstream project's figures left out]
3.
counted as 85/85 100/100 what do these even mean i just clicekd recommended last prompt
(Claude asked again, in plain words, how to report two areas' figures)
the recommended option for both
2026-10-28, commit 6807241, model claude-opus-5-5, effort xhigh.
1.
[a question about a game map of a workplace left out]
2.
[a follow-up about building plans for that map left out]
3. (answer to a next-step box question)
Yes, look it up
(Asked whether to retry with the VPN off after a DNS failure)
i dont have vpn on. always check if my vpn is even on before asking that. also always try builtin browser if curling etc fails
2026-10-28, commit 4b4ca60, model claude-opus-5-5, effort xhigh.
1. Quoted under 27b102b.
2. Quoted under 27b102b.
3. Quoted under 55fb3db.
4. Quoted under 8f63649.
5. Quoted under 55fb3db.
6. Quoted under b64c518.
7.
[an ask about a downstream project left out]
8.
is there a bug in my dotclaude as in why didnt [a teammate's] dotclaude close those off too when merging
9. (the last part answers whether to add the parent issue rule)
in the future also give a prompt i can give to [a teammate] so that i can "patch" their claude too without them having to fetch any changes themselves. yes add it
2026-10-27, commit 2390716, model claude-opus-5-5, effort xhigh.
10.
if the slack text is something that will survive a copypaste without your copyrich skill just paste it here in the future
2026-10-27, commit 8bfaf66, model claude-opus-5-5, effort xhigh.
9.
[an ask about the draft's content left out] dont use oy in non ext [a downstream project] slacks. [a wording fix in the draft left out] use double enters when newline. dont use inline links outisde [a downstream project] ext slack. [the rest, about the draft and colleagues, left out]
2026-10-27, commit f3c36cb, model claude-opus-5-5, effort xhigh.
8. (answer to whether the rename and the account are done)
i renamed to zululima, made the lambdakilo account, created lambdakilo.github.io there and added zululima as collaborator. do the rest
2026-10-27, commit 3dab69a, model claude-opus-5-5, effort xhigh.
17.
yeah leave it but in the future are you able to de-folder it in the first prompt if i accidentally started this in [a project]
18. (reply button)
add a rule that the first reply flags a session started in a folder its prompt has nothing to do with
2026-10-27, commit c443c34, model claude-opus-5-5, effort xhigh.
14.
can you move this from [a project] to this computer
(which folder to move the session to)
Home folder (Recommended)
15.
wot now its [the home folder's name] and This computer separately
(whether to make a scratch folder by hand)
should these no folder sessions have a folder or how should this be done?
16. (reply button)
make the scratch folder and move this session there
2026-10-27, commit 35517c5, model claude-opus-5-5, effort xhigh.
2. (answer to whether to fix the dream prompt)
yes fix the dream prompt that still says setup goes in the readme
2026-10-27, commit 9884fae, model claude-opus-5-5, effort xhigh.
1. Quoted under d265fb0.
2026-10-27, commit d265fb0, model claude-opus-5-5, effort xhigh.
1.
dotclaude compact the readme to a screenful like my other repos
2026-10-27, commit 12080eb, model claude-opus-5-5, effort xhigh.
9. (with a screenshot of the end of a reply)
[an answer about a downstream project left out] btw im never reading anything above where youve done something as i assume always that the thing i read as your summary and not anything above it.
2026-10-27, commit 1b5b94d, model claude-opus-5-5, effort xhigh.
1.
[a question about a study programme left out]
2.
[an ask for stickers left out]
3.
[an ask for another sticker left out]
4.
[an ask to rebrand the stickers left out]
5.
[an ask to follow a colleague's review left out]
6.
[a fix to the mascot artwork left out]
7.
[a clarification of that fix left out]
8.
[an ask to fix review comments left out]
9.
[the same ask again left out]
10.
[a question about a layout fix left out]
11.
[an ask for a comic sticker and its sources left out]
12.
[an ask for more comic stickers and a sticker pack left out]
13.
[pasted: an address] [a note on which source to use left out]
14.
[changes to the stickers and a second pack left out]
15.
[an ask for one more sticker left out]
16.
[a change to the stickers' text left out]
17.
bro just put them to one place pls dont make a mess of my downloads in the future
2026-10-26, commit 5c5695d, model claude-opus-5-5, effort xhigh.
1.
[an ask to draft a Slack reply for a downstream project left out]
2.
[a change to that draft left out]
3.
[a question about that project's work order left out]
4.
[an ask to draft the next reply left out]
5.
dont give me the slack link in the future. lipcaord thingy was good. [a research question about that project's test setup left out]
2026-10-26, commit b2b6cae, model claude-opus-5-5, effort xhigh.
1.
[an ask about a downstream project's issue comment left out]
2.
never draft replies to github issues. just explain here
2026-10-26, commit 8665fe4, model claude-opus-5-5, effort xhigh.
1.
[a question about a client contract left out]
2.
[quoted: a paragraph of Claude's earlier reply about the shared notes]
explain
3. (answer to whether to stop sharing those notes)
stop sharing them and you can force push to rebase and remove the things that would break contract
2026-10-26, commit edc5916, model claude-opus-5-5, effort xhigh.
1.
who opened my google chrome and why?
2.
cant you just always use builtin chrome
(How far should the rule reach?)
Only the Chrome app (Recommended)
2026-10-26, commit 1505c80, model claude-opus-5-5, effort xhigh.
26.
moving from step 3 to step 4 which is btw not correct it says step 1 hides the report part text fix
2026-10-26, commit d915bc3, model claude-opus-5-5, effort xhigh.
25. (with a screenshot of the review page)
hmmm i can see the next step too a bit. is there some css trick that could make me see only the current step in my screen?
2026-10-26, commit 27c6125, model claude-opus-5-5, effort xhigh.
24. (with a screenshot of the review page)
its not completely top theres still some leftover from last step in top
2026-10-26, commit 9f72aad, model claude-opus-5-5, effort xhigh.
23.
pressing x should move on to next step
2026-10-26, commit fec4f8f, model claude-opus-5-5, effort xhigh.
22.
oof when going to next step put it to the top since im using a laptop
2026-10-26, commit 3a61517, model claude-opus-5-5, effort xhigh.
21.
the n next open is gay. also
[pasted: a log of a setup command failing outside the checkout]
what happened with the subshell cd that weve been doing all the time? also remove the review panel on left. honestly not needed to be frozen. just tell me before steps if somethings not right or if findings. also i had to press j to go to the first and only then c worked. just goto first in the start. also steps taken from prs how to verify is false since i always generate my own steps based on also the prs steps. reword
2026-10-26, commit f4b9b3c, model claude-opus-5-5, effort xhigh.
20. (with a screenshot of the review page)
its still not centered. you can move the left side bar a lot to left and or you can also put half of it to right. also wth? step 1 already fixed then why you show me it? also what is manual steps claims? also change the browser title something that i can recognize immeditatley what it is instead of just general [a project's name] and it fits inside the tab size. also the link was way up in the convo this session put lower so i can just click
2026-10-26, commit cb18719, model claude-opus-5-5, effort xhigh.
19.
ok so. i want the current step centered. no need for the big ass texts like these [pasted: a check's detail line from the review page]
also its usually also the pr steps missed should be inline in the steps. if theres a command i should be able to copy it with c and it should tell me that i can use c to copy it. also the j and k to move down or up. findings should be first step so that i can still request changes first before i start a huge operation of manual checking.
? keys is also not too intuitive. just put the shortcuts there like press or x for check in the steps. remove teh ? keys
2026-10-26, commit 4bfacaa, model claude-opus-5-5, effort xhigh.
1.
can you alias s to gm instead the bin sod that is, you can actually alias the gm and rename the script start or day update and restart etc sh. also can it sleep for an appropriate while and restart my mac gracefully when updates are done
2. (answer to whether to update the README, a reply button)
yes, update the start of day section in the dotclaude readme to match and push
3.
i dont care to read the brew output i just want the system to restart gracefully cuz on debian if i instarestrat the packages are not yet updated completely
2026-10-25, commit 7073814, model claude-opus-5-5, effort xhigh.
3.
[pasted: a README sentence on what Claude Code reads] it reads much more and you know that. remove the line. does there exist a url scheme that git clones and opens in claude code? whats the optimal workflow for claude desktop code to get a new project underway? do i even have to clone this thing myself?
(What should the README's last line say?)
yeah can you actually give a copyable prompt to claude code? or does claude code desktop support a uri scheme that the person just opens to do that
2026-10-25, commit 139bc5e, model claude-opus-5-5, effort xhigh.
3. (answer to whether to fix the masking too)
also fix prompt-log.sh so it masks /home paths, here
2026-10-25, commit 3f8f767, model claude-opus-5-5, effort xhigh.
1.
[a question about model sizes left out] also fetch my whole ass dotclaude fromgithub
2. (answer to whether to fix the PROMPTS.md case flip on Linux here)
fix the prompts.md case flip on linux here
2026-10-25, commit 590df15, model claude-opus-5-5, effort xhigh.
1.
do you remember to still pause and tell me if my proton vpn is the reason you cant fetch some information? can you toggle it off and on later when that happens?
2026-10-25, commit 7425554, model claude-opus-5-5, effort xhigh.
38.
[a question about an IDE sign-in left out]
39.
[a feature request for an app left out]
40.
[an ask for research left out] also you didnt mention that anymore once i just gave you input. fix in the future
2026-10-25, commit 6a0a0da, model claude-opus-5-5, effort xhigh.
1.
readme is not short rn. make it so that the readmovfile link shows literally all of readme. [an ask for a diagram of a downstream project left out] newbies can just ask from claude how to start and develop
(Every project, or the one project only?)
Every project (Recommended)
2026-10-25, commit 05a61d8, model claude-opus-5-5, effort xhigh.
36.
[an ask about an app's log left out]
(a message sent while that prompt ran)
[pasted: a screenshot of a simulator's crash report dialog]
2026-10-25, commit 4ad0136, model claude-opus-5-5, effort xhigh.
31.
[a question about a store review left out]
32.
[a store review check left out]
33.
[a choice about a store review left out]
34.
[a remark about a tester invitation left out]
35.
[a tester sign-up check left out]
36.
lol dont warn me about the dev mode. [an ask about an app's log left out]
2026-10-25, commit e9533e7, model claude-opus-5-5, effort xhigh.
11.
my guy. if you need something from me from the pane in the future also open the the pane for me. or is that possibel? [a note that a sign-in was done left out]
2026-10-25, commit 5aa28d3, model claude-opus-5-5, effort xhigh.
1.
remove dollars and token amounts from taskbar but keep them in the last dream report of the week
(Should the eve, night and day reports drop them too?)
Final report only (Recommended)
2026-10-25, commit ae13644, model claude-opus-5-5, effort xhigh.
30.
can you close the browser and simulation windows when theyre done or can claude even do that
2026-10-25, commit ad4d6ba, model claude-opus-5-5, effort xhigh.
24.
[a feature request for an app left out]
25.
2 [app] ended suddenly apple messages came to my screen
26.
third one
27.
twice again pls stop its interrupting my other work
28.
[a question about the crash left out]
29.
[a question about a store listing left out]
2026-10-24, commit 938ef43, model claude-opus-5-5, effort xhigh.
5.
[questions doubting the service's privacy and data residency claims left out]
6. (answer to whether to widen the rule)
yes, widen the new CLAUDE.md rule so a service's claims are checked against its own docs and privacy policy before a reply repeats them
2026-10-24, commit ed9a4e7, model claude-opus-5-5, effort xhigh.
1.
[a question about a hosted agent service left out]
2.
[a question about the same service for colleagues left out]
3.
my guy my setup can install git, jq and gh there and my dotclaude connects to slack, drive, email etc
4. (answer to whether to add the rule)
yes, add a rule to CLAUDE.md that a session reads a setup's own install scripts and docs before saying it can't do something
2026-10-24, commit 7bed059, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2. (asking why a dream step was put to the user) Quoted under 12dd424.
2026-10-24, commit b1b0ad5, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2. (asking why a dream step was put to the user) Quoted under 12dd424.
2026-10-24, commit d048675, model claude-opus-5-5, effort xhigh.
3. (asking why the weekly usage stood at 67% near the reset)
holdup what the weekly usage is reseting soon but i see only 67 percent used
4.
you can also tell me honestly next time when youre done dreaming that you had nothing else to do before usage could have gone to max
2026-10-24, commit 16dae55, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2. (asking why a dream step was put to the user) Quoted under 12dd424.
2026-10-24, commit fbb41ac, model claude-opus-5-5, effort xhigh.
1.
dream
(Add p9, hand the dream review to a new session?)
Add it (Recommended)
2. (asking why a dream step was put to the user) Quoted under 12dd424.
2026-10-24, commit 5ea9c06, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2. (asking why a dream step was put to the user) Quoted under 12dd424.
2026-10-24, commit 12dd424, model claude-opus-5-5, effort xhigh.
2. (asking why a dream step was put to the user)
wait why you ask it was not shady nor behavior
2026-10-24, commit 3acfa5c, model claude-opus-5-5, effort xhigh.
1.
dream
(Add p1, sync.sh --commit takes only the paths named?)
Add it (Recommended)
2026-10-24, commit b6a04c3, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit f7193a3, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit 0fb383a, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit a4494ba, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit bc7da4d, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit 1e5097f, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit 2baa204, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit f144ac4, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit ee853ef, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit bc7e15a, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit 3e67088, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit 1cb91bc, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit 34ec29d, model claude-opus-5-5, effort xhigh.
1. Quoted under d509ffb.
2026-10-24, commit d509ffb, model claude-opus-5-5, effort xhigh.
1.
dream
2026-10-24, commit cbd7e8e, model claude-opus-5-5, effort xhigh.
18.
theres still two pr drafts on my end open on [a project's] frontend fix to the origianl pr
2026-10-24, commit 3cab280, model claude-opus-5-5, effort xhigh.
10.
Do this task here: "Carry two new wiki edits into the intra handbook"
(How should the two wiki edits, a colleague's contact row and the new expense app paragraph, land in the handbook repository?)
hold the phone what. [a colleague's] contact row? they should add it themselves. fix in future so that you dont suggest even edit like this
2026-10-24, commit 3036182, model claude-opus-5-5, effort xhigh.
17.
bro what. just incorporate always the things to the same pr
2026-10-24, commit c4bd0fe, model claude-opus-5-5, effort xhigh.
16.
bruh now its hidden in more. also the header takes a lot of screen space by being frozen there. you can put frozen stuff to the sides maybe?
2026-10-24, commit 4bef872, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 7f326bb, model claude-opus-5-5, effort xhigh.
15.
[a screenshot of the review page open in a browser] pretty good but j doesnt go to first step. also the first chapter is maybe a bit too long as you can see from the ss. also localhost should be a clickable link on the html
2026-10-24, commit 40bb88c, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit c6d5bb8, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 1d2e084, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 60f3bf1, model claude-opus-5-5, effort xhigh.
9.
ok these are pretty esoteric. in the future just ask me a input if its something shady/security/privacy or if i have to change my behavior so i ack it
2026-10-24, commit d1a5e8a, model claude-opus-5-5, effort xhigh.
4.
dream review
(p33: hold the fenced-command rule to Markdown a person reads, add it?)
Add it (Recommended)
2026-10-24, commit 871f637, model claude-opus-5-5, effort xhigh.
4.
dream review
(p32: shrink the two stack-hook bullets to one and require a README entry for each new hook, add it?)
Add it (Recommended)
2026-10-24, commit 3d4cb0f, model claude-opus-5-5, effort xhigh.
13.
nah bro no artifacts create a local html in appropriate place and give me link i can click instaed of copy
2026-10-24, commit 6d60275, model claude-opus-5-5, effort xhigh.
4.
dream review
(p30: move the mirroring mechanics from CLAUDE.md into a dotclaude skill, add it?)
Add it (Recommended)
2026-10-24, commit 1d9861a, model claude-opus-5-5, effort xhigh.
4.
dream review
(p19: set autoCompactWindow to 400000 in settings.json, add it?)
Add it (Recommended)
2026-10-24, commit 49d562d, model claude-opus-5-5, effort xhigh.
4.
dream review
(p14: commit the fixes a check on the user's own PR recommends, add it to dotclaude?)
Add it (Recommended)
2026-10-24, commit c163935, model claude-opus-5-5, effort xhigh.
8.
oof can you add a p12 out of something so ican see how manyproposals still left
2026-10-24, commit a984b3e, model claude-opus-5-5, effort xhigh.
4.
dream review
(p12: add shwordsplit to the Claude-only zshrc block and a CLAUDE.md bullet on zsh, add it?)
Add it (Recommended)
2026-10-24, commit a74e0d8, model claude-opus-5-5, effort xhigh.
11.
lets start doing a new way to review prs. theres a mismatch between github prs and the notes here. instead always generate a html that i can open as link here that has steps that track what steps ive already done with sane keyboard shortcuts. it should also have the screenshots from github if any
2026-10-24, commit 81f40fa, model claude-opus-5-5, effort xhigh.
4.
dream review
(p9: have sync.sh parse every mirrored script, require README entries and cap prose at 100 columns, add it?)
Add it (Recommended)
2026-10-24, commit d72016d, model claude-opus-5-5, effort xhigh.
4.
dream review
(p7: deny seven never-called tools in settings.json, add it to dotclaude?)
Add it (Recommended)
2026-10-24, commit 4fccd4e, model claude-opus-5-5, effort xhigh.
4. Quoted under 01f7bfc.
2026-10-24, commit 7c27206, model claude-opus-5-5, effort xhigh.
4. Quoted under 01f7bfc.
2026-10-24, commit 01f7bfc, model claude-opus-5-5, effort xhigh.
4.
dream review
(p6: replace the never-run-remote-tooling bullet with one that allows only an asked-for action and target, add it to dotclaude?)
add it but that was way too abstract. remote tooling could mean anything. you must give always an example from history
2026-10-24, commit f709146, model claude-opus-5-5, effort xhigh.
4. Quoted under a18fc01.
2026-10-24, commit debda14, model claude-opus-5-5, effort xhigh.
4. Quoted under a18fc01.
2026-10-24, commit a18fc01, model claude-opus-5-5, effort xhigh.
4.
dream review
(p5: put the closing reply after the box and check Prompt feedback on Stop, add it to dotclaude?)
add it but ok so in the future assume that i dont know anymore during review what different hooks do so you must explain everything a bit more from the beginning like what happens now and what would happen after adding the proposal
2026-10-24, commit f76c0d0, model claude-opus-5-5, effort xhigh.
7.
whats the current "prompt" that writes my pr descriptions currently. compare those
8.
lets borrow things from [a colleague's] skill to my dotclaude [and an ask about a pull request left out]
(where the list of changes goes)
As drafted (Recommended)
2026-10-24, commit 0b234e8, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit e2fdbc4, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 9786c64, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 9f9a104, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 9175bcf, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit cf41759, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 83f998c, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 87e7c9e, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 4bf2aaf, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 78f55fa, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit d3d65e8, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit f5a83f9, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 8240afe, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit fbb71b8, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 47bc411, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 856554b, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit f74e173, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 7ca211e, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit fb0545a, model claude-opus-5-5, effort xhigh.
5.
dotclaude always tell me and do the ask dialogue if you have a hunch my proton vpn is blocking your requests somehow
2026-10-24, commit 1c85586, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 25fa6f5, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 9b94b7b, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit 8d80617, model claude-opus-5-5, effort xhigh.
4. Quoted under eee020a.
2026-10-24, commit eee020a, model claude-opus-5-5, effort xhigh.
4.
dream review
2026-10-24, commit 9e3cc9e, model claude-opus-5-5, effort xhigh.
3. Quoted under 5c100c9.
2026-10-23, commit fb29876, model claude-opus-5-5, effort xhigh.
1.
[pasted: the prompts section of 5c100c9, with the questions and picks of a dream review on nested lines]
why do you add these double > it looks a bit confusing
(How should AskUserQuestion answers appear under a quoted prompt?)
Paren + quote (Recommended)
2026-10-23, commit 32a32d7, model claude-opus-5-5, effort xhigh.
2. Quoted under 5c100c9.
3. Quoted under 5c100c9.
2026-10-23, commit 5c100c9, model claude-opus-5-5, effort xhigh.
1.
dream review
p16 (medium): prompt-log.sh prints the prompts heading with model and effort itself, and exits 1 instead of printing nothing. Add it to dotclaude (hook plus the printing bullet in CLAUDE.md)? Add it (Recommended) p17 (medium): ship the browser-pane picture upload as pane-upload.sh beside the pull-requests skill, and point the skill at it. Add it to dotclaude? Add it (Recommended) p18 (medium): keep only pronouns and [the answer preferences] in rules/about-me.md (loaded everywhere), move the rest unchanged to ~/.claude/about-me/profile.md (read on demand), and add two lines to CLAUDE.md's Memory hygiene. Add it (profile stays local, CLAUDE.md lines go to dotclaude)? Add it (Recommended) p19 (medium): a No folder session asked to start a lasting repo creates ~/Projects/<name> itself and moves there before the first file. Add it to dotclaude's CLAUDE.md? Add it (Recommended) p20 (medium): widen the AskUserQuestion bullet so an assumed or contradicted public fact about a person (affiliation, supervisor, employer) is asked on its own, never left as a bullet in a longer reply. Add it to dotclaude's CLAUDE.md? Add it (Recommended)
2.
is it still so that you cant auto incorporate these proposals during dream review in auto mode?
How should the remaining 42 dream steps (p21 to p62) go? Auto-apply, ask exceptions (Recommended)
3.
heading to sleep. start dreaming after youve applied all auto things
2026-10-23, commit 8e1a04e, model claude-opus-5-5, effort xhigh.
4.
what? you gave me nothing to copy
2026-10-23, commit 9a5644c, model claude-opus-5-5, effort xhigh.
1.
[an ask to rewrite a note left out]
2.
[an ask about the same note left out]
3.
the word stretch already is half cut horizontally
2026-10-23, commit 038b327, model claude-opus-5-5, effort xhigh.
1.
[an ask about a Slack draft for work left out]
2.
[an ask about the same draft left out]
3.
[an ask about the same draft left out]
4.
[an ask about the same draft left out]
5.
for slack messages never give me the copyable thingy just give me only the command to copy like on the other session
2026-10-23, commit ddc0bb5, model claude-opus-5-5, effort xhigh.
10.
[a shell command the user ran left out]
11.
okei jatkossa älä edes anna jatkossa tuota kopioitavaa juttua suoraan vaan toi copyrich komento
2026-10-23, commit b17576c, model claude-opus-5-5, effort xhigh.
5.
[a shell command the user ran left out]
6. Quoted under b64c518.
7.
[a correction about a downstream project's channel left out]
8.
[a correction about a downstream project's channel left out]
9.
pistä vaan jatkossa esim. commit messaget linkattuna inline. voit myös käyttää code format koodi juituille esim. [a table name]. esim. [a page of a downstream project's old app] vois olla kans linkki sinne
2026-10-23, commit 40bd746, model claude-opus-5-5, effort xhigh.
1. Quoted under 27b102b.
2. Quoted under 27b102b.
3.
[a question about opening a downstream project's file left out]
4.
dotclaude if you ever do something that literally pops open a window in my screen ask first
2026-10-23, commit f38a703, model claude-opus-5-5, effort xhigh.
2.
maybe it should be on readme instead
2026-10-23, commit a2cad92, model claude-opus-5-5, effort xhigh.
1.
[a research ask about a university course left out]
2.
[a personal fact to remember left out]
3.
[an ask for a chat message draft left out]
4.
did you check the last signal session about how many lines it can actually have before wrapping? save for future as well the char amount
2026-10-23, commit 3a5aff5, model claude-opus-5-5, effort xhigh.
1.
dotclaude i should add a how i work .md file there as well that explains how i work with dotlcaude i.e. nothing that gives away customer things like ip but how i start my day with bin/sod, how i prompt and use claude desktop, firefox and the settings my m5 air uses to prevent 80% battery over. it should also say the how i do first dreaming on friday when i get home and hook up charge. then more work on saturday and then charge and dream rest.
2026-10-22, commit 539cb34, model claude-opus-5-5, effort xhigh.
1.
is it even possible to do any evals on dotclaude?
2.
even parts of it?
3.
white hook tests also could we publish some evals when done during dreaming to the readme.md? update them as well?
4.
ask the questions again
Rule evals: Yes, counts only (Recommended) Update: Refresh results each run (Recommended)
6. (answer to Claude's question, a reply button)
both scripts ran, finish it
2026-10-22, commit 6a5818e, model claude-opus-5-5, effort xhigh.
1.
dreaming should also include [a shared AI library], [a CLI] and [a project template] if theres anything we should add to the projects that derive from those like [a downstream project].
2026-10-22, commit 95ed02e, model claude-opus-5-5, effort xhigh.
7.
kirjakieli also only in [a downstream channel]
8.
[the user's edit of a draft about a downstream project left out]
when just sending messages to non ext [a downstream project] slack you dont have to do the full company names. also the joo at the beginning was not very finnish of you
2026-10-22, commit 55fb3db, model claude-opus-5-5, effort xhigh.
3.
[an ask about a downstream project left out]
4. Quoted under 8f63649.
5.
[an ask about a downstream project left out]
6.
[pasted: a colleague's two questions about progress figures] [an ask about a downstream project left out] also dotclaude only use the copypaste thingy for [a downstream channel]. also dont open the builtin web brwoser whe doing that
2026-10-22, commit ffd6ec3, model claude-opus-5-5, effort xhigh.
1. Quoted under 27b102b.
2.
what. in future i wanna see the same format as asked in your replies:
[pasted: a colleague's nested list of progress figures, 12 lines]
like this
2026-10-22, commit 83f4ebf, model claude-opus-5-5, effort xhigh.
29.
again with the things being in other replies. dotclaude fix. next time when you give instructions (give again) add the qr for example so that each reply is complete without the other replies being read
2026-10-22, commit 9e28119, model claude-opus-5-5, effort xhigh.
1. Quoted under 27b102b.
2.
[a question about a draft for a downstream project left out]
3.
dont save html temp files like that in my Downloads store them in .claude etc somewhre where it doesnt delete like tmp but not something that syncs
2026-10-22, commit 3ebb7db, model claude-opus-5-5, effort xhigh.
1.
make the description oneliner. also topics only oneliner. use builtin brwoser to check newline length. also add somehwere that my i think this is the next step in free software / open source because say they use ai but give only source code back, never how they use ai. i say put your money where your mouth is.
2026-10-22, commit db33bc4, model claude-opus-5-5, effort xhigh.
1.
can claude code still not rename sessions here?
2.
can we try to rename this session to foobar
3.
can you rename it to something random now
4.
change this session to numbers going without spaces to 30
5.
dotclaude after each prompt rename my sessions accordingly based on the session so far. you can see from the screenshot when the titel on my m5 air fades and cuts out [pasted: a screenshot of the session sidebar]
2026-10-22, commit 8b750ae, model claude-opus-5-5, effort xhigh.
6.
what is that dead weight rule? also explain that new session suggestion you gave me
7.
y
2026-10-22, commit d7e83e0, model claude-opus-5-5, effort xhigh.
1.
can you remove src even though its just symlink and removing it doesnt break anything
Fix hooks, remove later (Recommended)
2. (answer to Claude's question, a reply button)
fix the two hooks here, then sync and commit
2026-10-22, commit cfdf0b3, model claude-opus-5-5, effort xhigh.
2.
[an ask to push a downstream project's work straight to its default branch left out]
3.
you didnt give me no allow script above
4.
ran
5.
ran
2026-10-22, commit 8c0bcdc, model claude-opus-5-5, effort xhigh.
1.
could we get human readable readme in dotclaude? its very long, too verbose. try to write it to a vibe coder who actually doesnt understand how code works
2. (answering whether to commit and push the new README)
commit and push the new readme
2026-10-22, commit bfc2d36, model claude-opus-5-5, effort xhigh.
2. (answering whether sync.sh should list the PRs pinned to an older commit)
yes, make sync.sh --commit list my open prs that carry the dotclaude block but pin an older commit, right after it pushes
2026-10-22, commit 4f8fc53, model claude-opus-5-5, effort xhigh.
1.
dotclaude when i make edits to dotclaude remember to check all my open pull requests to match for example do have link to the commit at the time of dotclaude etc. understand?
2026-10-22, commit 92d1e9c, model claude-opus-5-5, effort xhigh.
3. (answer: the user ran Claude's sed edit of the hook, which auto mode had refused to make)
ran
2026-10-22, commit 0b9780d, model claude-opus-5-5, effort xhigh.
1.
[an ask about a fast-forward error in another repository left out]
2. (answer to Claude's question, a reply button)
yes, [two changes to another repository left out], and fix the same ff-only pull in sync.sh's mirror step
2026-10-22, commit 49b8a20, model claude-opus-5-5, effort xhigh.
1.
when i rebase something i force push as in rewrite history does git keep the original date they was committed at?
2.
does github show author date or committer date
3.
yeah put so that rebase date keeps the original date in my dotclaude
2026-10-22, commit 16d04eb, model claude-opus-5-5, effort xhigh.
1.
[an ask about a work task left out]
2. (answer to an offer to draft a Slack post)
always draft a slack message the way you did it last time in some session? maybe a command to copypaste it to clipboard?
2026-10-22, commit 7102d15, model claude-opus-5-5, effort xhigh.
1.
[a question about a downstream project left out]
2.
[a question about a downstream project left out]
3. Quoted under c558028.
4.
[a question about a downstream project left out]
5.
[a question about a downstream project left out]
6.
how did you come up with 5% rest hookup percent
2026-10-22, commit df95b19, model claude-opus-5-5, effort xhigh.
28.
dotclaude you instructions were in wrong order. you first told me to record then the instrructions to install app to phone. [a note about a project left out]
2026-10-21, commit a174303, model claude-opus-5-5, effort xhigh.
20. Quoted under 1538093.
23. Quoted under 1538093.
2026-10-21, commit 1538093, model claude-opus-5-5, effort xhigh.
20.
also can you make dreaming also analyze the [shared repository] in its entirety? trimming unnecessary memories sounds like a good way to spend tokens
23.
lets no start dreaming now ill do it on friday and saturday but just include it in dreaming then
2026-10-21, commit 0a5b6b1, model claude-opus-5-5, effort xhigh.
7.
put these to the appropriate context (when i say in future context you know where to put them memory, project memroy global ruling, project docs etc, remember in dotclaude)
[three rules for a downstream project left out]
ask questions if unclear you can push these additions to the prompts.md pull request
[four answers about those rules left out]
2026-10-21, commit 05ac304, model claude-opus-5-5, effort xhigh.
15. Quoted under 9294539.
2026-10-21, commit 9294539, model claude-opus-5-5, effort xhigh.
13.
can i just poke a hole in the automode so it can exfil info to memories dir in [a shared repository]
14.
cant i have a memory sharing script / hook for sharing memories to appropriate places in the [shared repository]/memories?
15.
it shouldnt be opt-in it should be automatic. yes dotclaude set up the hook
16. (a reply button's words)
switch this session to ask before each write and build the hook, i approve each step
17.
its now on manual. the hook should work on automode right after we build it?
18.
also dotclaude since you clearly couldnt set the mode yourself never give me a suggestion to say to you switch modes just say to me i have to do it myself
19.
should memories dir hold just everyones memories in there or separate by user like lambdakilo there
2026-10-21, commit b64c518, model claude-opus-5-5, effort xhigh.
6.
[an ask about a downstream project left out]
7.
[a question about a downstream project left out]
8.
format it so that it uses slack formatting to its advantage. for example commit message could be a link to the repo at the commit time etc
9.
[a shell command the user ran to copy the draft to the clipboard]
10.
it doesnt give me anything in clipboard. [an ask about the draft's wording left out]
11.
now it got to my clipboard. save this feature to dotclaude
2026-10-21, commit c558028, model claude-opus-5-5, effort xhigh.
2. Quoted under 27b102b.
3.
[a question about a downstream project left out]
4. Quoted under 8f63649.
5.
did you check the [a deployed test environment], [a legacy app] and every other connector and context before you drafted that
2026-10-21, commit 319428f, model claude-opus-5-5, effort xhigh.
1.
[an ask about a downstream project's chat thread left out] never send messages on behalf of me on slack
2026-10-20, commit daa423d, model claude-opus-5-5, effort xhigh.
6.
your example was not clear and concrete enough
p15 (medium): refuse whole-disk and home-wide searches with a PreToolUse hook in every session. Add it? Add it (Recommended)
2026-10-20, commit 99a557c, model claude-opus-5-5, effort xhigh.
6.
your example was not clear and concrete enough
p14 (medium), trimmed: stop Prompt feedback calling a dotclaude ask unclear or a second job. Add it? Add it (Recommended)
2026-10-20, commit 3450279, model claude-opus-5-5, effort xhigh.
5.
your example was not clear and concrete enough
2026-10-20, commit 3216b16, model claude-opus-5-5, effort xhigh.
4.
give an example of p12 then ask again from me
p12 (high): run the dream workflow's agents as a custom agent type that carries only the tools they use. Add it? Add it (Recommended)
2026-10-20, commit 992d9b7, model claude-opus-5-5, effort xhigh.
12.
you didnt tell me the issue names. fix in the future and now
2026-10-20, commit 7684d53, model claude-opus-5-5, effort xhigh.
4.
give an example of p12 then ask again from me
2026-10-20, commit bf29ab3, model claude-opus-5-5, effort xhigh.
11.
also dont say dot in the filename if not necessary for example 20-oct would have been better
2026-10-20, commit d1a6ed5, model claude-opus-5-5, effort xhigh.
3.
what does w mean here? also if you dont have any examples stored anywhere invent some? should i just say p1 like proposals
2026-10-20, commit 7c32637, model claude-opus-5-5, effort xhigh.
1.
dream review
2.
what does 43 more pending next run mean? also you didnt give concrete examples of the w5p1. fix in the future and your reply.
2026-10-20, commit bf06b4e, model claude-opus-5-5, effort xhigh.
9.
also never use underscores in the filenames you can use dashes -
2026-10-20, commit c0a94a8, model claude-opus-5-5, effort xhigh.
1.
dotclaude maybe add in my prs instead of link to my dotclaude add link to the repo in a commit time when the pr was last edited
2026-10-20, commit 6fe58fc, model claude-opus-5-5, effort xhigh.
4.
restart the session so i see it
5.
can the session usage use emojis or colors lol like cold for ice and fire for warm?
2026-10-20, commit 34e7c44, model claude-opus-5-5, effort xhigh.
3.
is the text session-usage necessary in statusline?
Rename the mod so its label in the status line is shorter? usage (Recommended)
2026-10-20, commit e619ae3, model claude-opus-5-5, effort xhigh.
4.
if somethings medium or high (anything not low) give me shortest possible reproduction steps to achieve the bug so that i can just tell the author to fix it first before i review everything visually
2026-10-20, commit 363d13f, model claude-opus-5-5, effort xhigh.
2.
[pasted: a screenshot of the status line cut off after the cost] nope thats too long it wraps already in status line
The app shows about 40 characters. Which status line layout should the mod use? Cache first, short (Recommended)
2026-10-20, commit c27220d, model claude-opus-5-5, effort xhigh.
1.
add to the status line warm cache until num Month then the time because after a day i cant say anymore what day the cache was. is the status line dynamic? can it show like how many minutes until cache expire or if already expired?
2026-10-20, commit 53bed44, model claude-opus-5-5, effort xhigh.
16.
dont add p9 and also add co2 and kwh at the end of dreaming. add also to dream workflow a report that i can see on monday on how many tokens were used, dollars would have cost, co2 and kwh of the whole weeks estimated as an aggregate of all sessions. session should somehow store these metrics in the computer in each prompt because server location and time of day matter for kwh. also add a metric of how many tokens (percentage too) hit the warm cache
2026-10-20, commit 23969ef, model claude-opus-5-5, effort xhigh.
3.
[a note that a review step named other test data than the PR's screenshot showed] fix this now and in future
2026-10-20, commit 4368f4d, model claude-opus-5-5, effort xhigh.
3.
always delete merged branches
2026-10-20, commit 428a710, model claude-opus-5-5, effort xhigh.
1.
[an ask to merge a pull request in a downstream project left out]
2. (after running the script that adds the two rules)
ran it, go on
2026-10-20, commit 648b928, model claude-opus-5-5, effort xhigh.
1.
dotclaude add estimates of kilowatthours and co2 grams to the statusline with context window. those could come from estimates of time and datacenter area. the statusline is separate operation of the session context window right?
2026-10-20, commit c224c07, model claude-opus-5-5, effort xhigh.
15.
after dreaming is done does it say how many tokens and usd were spent? if not, fix
2026-10-20, commit 9d08ba7, model claude-opus-5-5, effort xhigh.
1.
[a review of a pull request in a downstream project left out]
2.
always say in the review steps where a bug can be seen like the greyed out bug now i dont know where to see it in the steps
2026-10-20, commit 8aca51a, model claude-opus-5-5, effort xhigh.
14. Quoted under 15d672a.
2026-10-20, commit 15d672a, model claude-opus-5-5, effort xhigh.
14.
ok but how do i continue working with the previous dream results now?
2026-10-20, commit bef9337, model claude-opus-5-5, effort xhigh.
13.
theres a lot of proposals after every dream. let alone after fri and sat on mon. lets do it so that after every dream the session asks in dialogue if i wanna add it to dotclaude, not or something else. the dialogue should give me the exact proposal on what is added and and a concrete example on what is wrong and how the example looks if fixed
2026-10-20, commit 1732363, model claude-opus-5-5, effort xhigh.
4.
lets test the usage line
6.
what. why did it open in another window
2026-10-20, commit 0dbf3f0, model claude-opus-5-5, effort xhigh.
1.
[pasted: a screenshot of the usage line in a Claude Code notice] context window messages are now in claude code notice dont put them in there. also its not a oneliner on my m5air. shorten it so that it doesnt wrap
Where should the usage line go instead? App status line (Recommended)
3.
wth i did not decline that i pressed many times allow but i just reloaded
2026-10-20, commit 44b93de, model claude-opus-5-5, effort xhigh.
19. Quoted under 49e41eb.
2026-10-20, commit 49e41eb, model claude-opus-5-5, effort xhigh.
19.
also when you say copy something put that in a triple backquote whatever code block so i can one click copy it even if its an url address
[an ask about a downstream project left out]
2026-10-20, commit 74a5666, model claude-opus-5-5, effort xhigh.
16.
[an ask about a downstream project left out]
17.
[an ask about a downstream project left out]
18.
dotclaude command option j doesnt open dev tool console btw on firefox. [an ask about a downstream project left out] also edit dotclaude so that i dont have to say dotclaude everytime i wanna make a diff in global ruling. in the future infer that if i have a thing i want fixed in future prompts and sessions as well youll add it to dotclaude. ill say explicitly if i only want a change in this session or project or etc
also remove the dotcalude stale cache guard prompt. instead show me a oneline of how much context percentage and tokens and maybe dollars has been spent already on the session and when does the prompt cache expire
2026-10-19, commit e1d30be, model claude-opus-5-5, effort xhigh.
1.
[an ask to turn a book into audio left out]
2.
[an ask to narrow it to two chapters left out]
3.
dotclaude next time when you textify the pdf take more care to make sound breaks on sub chapters like "[a section heading]" was in the [a book]
4.
go, add the rule and push it to dotclaude
2026-10-19, commit 656a92e, model claude-opus-5-5, effort xhigh.
12.
its now monday and im looking at the session limit message. any stats what happened? i should be seeing the report now. at least my weekly usage is 0% now so thats good [pasted: a screenshot of the app's session limit message]
Which of the 51 tested dream scripts should I apply and commit now? do the high impacts first. but dotclaude ask me each proposition one by one in the future because i cant scroll up to p2 etc from p60 and still remember the order and numbers and your comments from down below. thoughts? p7 and p9 both add a hook against a cd that later commands don't depend on. Keep both? Only p7 (Recommended)
2026-10-19, commit 6c05d5d, model claude-opus-5-5, effort xhigh.
14.
[an ask about a downstream project left out]
15.
isnt this already in dotclaude? always give terminal commands so that i can run them from home dir instead. rename my src to Projects as per the new XDG upcoming standard
2026-10-19, commit 5df2771, model claude-opus-5-5, effort xhigh.
11.
dream
12.
its now monday and im looking at the session limit message. any stats what happened? i should be seeing the report now. at least my weekly usage is 0% now so thats good [pasted: a screenshot of the app's session limit message]
How should I recover Saturday's dream results? Fix, then resume (Recommended)
2026-10-17, commit 4feb84f, model claude-opus-5-5, effort xhigh.
3. Quoted under ff9d05b.
4. Quoted under ff9d05b.
2026-10-17, commit 992ac4c, model claude-opus-5-5, effort xhigh.
3. Quoted under ff9d05b.
4. Quoted under ff9d05b.
2026-10-17, commit ff9d05b, model claude-opus-5-5, effort xhigh.
3.
not bad. [a question about which repositories could use a PROMPTS.md left out]
4.
[an ask to add a PROMPTS.md to two projects left out] also sometimes you drop the effort level from prompts.md. dotcalude debug and fix
2026-10-17, commit 4f5472e, model claude-opus-5-5, effort xhigh.
1.
what sort of file name would be good for for example dotclaude prompts.md that would list all prompts in the repo, the relating commit to that and the pr? would it be PROMPTS.md or prompts.md?
2.
dotclaude always update a PROMPTS.md in a repo. push one to dotclaude repo and ill check if its so good so i can push it to my other repos
2026-10-17, commit 3a679ab, model claude-opus-5-5, effort xhigh.
25.
[a note about a project left out] dotclaude if not personal dont tell me to do inputs on browsers etc. for example dont list the eu trader status. if not sure, ask me. otherwise continue on your own after im done inputingg my personal information. [a note about a project left out]
26.
take the rule, go
2026-10-17, commit 64b0e44, model claude-opus-5-5, effort xhigh.
11.
[an ask about a downstream project left out]
12.
does claude always edit the pr description accordingly after fixing comments?
13.
go rule
2026-10-17, commit 7171229, model claude-opus-5-5, effort xhigh.
9.
dotclaude dont give me those instructions 2 and 3 since theyll anyway be lost in the dreaming output. if you want you can give them to me after dreaming has been done.
[an ask about a usage check after the dream run left out]
10. (answering whether to add the drafted sentence to the rule)
go, add the drafted sentence to the to-do list rule in CLAUDE.md
2026-10-17, commit cdcd62e, model claude-opus-5-5, effort xhigh.
9.
dotclaude dont give me those instructions 2 and 3 since theyll anyway be lost in the dreaming output. if you want you can give them to me after dreaming has been done.
also dotclaude can you give me a verification after saturday dreaming that 5h and weekly usage was spent to x percent after dreaming completed
2026-10-17, commit 4b564c1, model claude-opus-5-5, effort xhigh.
21.
dotclaude always give the commands like i could paste them from home folder, i.e. [a path inside a project] doesnt work here. [questions about a project left out]
22.
take the new wording, [a note about a project left out], go
2026-10-17, commit 489399d, model claude-opus-5-5, effort xhigh.
6.
[an ask about a weekly usage recap in the dream report left out] also dotclaude you told me stuff like "dont run p5 p1 and p4 are still yours to run. always give me a full list of what to do next since i shouldnt remember things like that on my human memory if you make changes to the todolist after dreaming changes on the sat morning
7. (answering whether to add the rule as drafted)
go, add the full to-do list rule to CLAUDE.md as drafted
8. (answering whether to commit p1 to p4 and the new rule as two commits) Quoted under 2250924.
2026-10-17, commit 2250924, model claude-opus-5-5, effort xhigh.
4. Quoted under 5d38f18.
8. (answering whether to commit p1 to p4 and the new rule as two commits)
yes, commit p1 to p4 first and then the new rule, as two commits, and push
2026-10-17, commit dfc134f, model claude-opus-5-5, effort xhigh.
7.
weekly usage history should be a nice recap in the chat in the end of dreaming like percentages last week, last month average, last year average. [an ask to always get the whole to-do list after a change to it left out]
2026-10-17, commit 5d38f18, model claude-opus-5-5, effort xhigh.
4.
run the dream workflow now
6. (answering whether to fix the four dream.js problems)
Yes, fix them. also my mac asked me apparently during dreaming can i read files permission. i only clicked accept probably wayyy into the dreaming. does this affect me? can i somehow give claude permissions forever reading my files? also does anthropic store my weekly usage at the end of weeks somewhere? if not can i start collecting them with dreaming to some place here in local since dreaming is the last thing before my weekly resets?
2026-10-16, commit bc05825, model claude-opus-5-5, effort xhigh.
3.
also have you put the documentation etc of claude dream etc [pasted: the address of the Managed Agents dreams page in the Claude Platform docs] into the things it can check and analyze if anthropic has invented something new that my dream workflow could use
2026-10-16, commit a999519, model claude-opus-5-5, effort xhigh.
1.
you said my weekly usage wont even fill up currently with my dreaming workflow. what else could my dream workflow analyze etc to make it fill up the rest of the usage? maybe also prioritize most important things before dreaming the less important things? like more recent sessions more important than the 90 day old session?
also can i somehow speed up the dreaming progress? weekly resets happen 17 on saturday and im still doing some work on saturday morning maybe around 8 or 9 and i dont think i can dream fast enough to fill weekly usage.
also should i dream even on friday evening to fill the 5 hour windows so that one more 5 hour usage can fill up the rest of the weekly usage when begun around 11?
2. (answering whether to build the ranked queue and new sources)
yes build it but also cant i just say dream and it checks if its on friday it can run the 5 hour windows and weekly so that on saturday i still have weekly usage for one 5h window and one actual 5 h fresh window for work, and the saying dream on saturday checks its saturday and spends the rest of the weekly and 5h usage there. hopefully using the whole 100% on 5h and weekly or stopping just before those so that the agents can return and finish the dreaming gracefully?
2026-10-16, commit c4657b7, model claude-opus-5-5, effort xhigh.
27.
regenerate p26 from the weekly report
28.
ran p26, commit and push
2026-10-16, commit 2360937, model claude-opus-5-5, effort xhigh.
26.
keep transcripts for 90 days
2026-10-16, commit e4194a7, model claude-opus-5-5, effort xhigh.
25.
could we do it so that a dream with weekend as parameter so it burns also through the weeks usage after 5 hour window resets
2026-10-16, commit 0b253b5, model claude-opus-5-5, effort xhigh.
24.
arent all dream workflows overnight? why do i need to say that again when doing just slash dream?
2026-10-16, commit 6f5bea1, model claude-opus-5-5, effort xhigh.
17.
i didnt understand the habit of p10 so i didnt run the bash command. explain and list the the rest of the items again from 10 onward.
also just like you saw the weekly didnt really eat that much tokens after all. could i use this even more frequently when having idle time like during a lunch break, end of day over night and over the weekend? would i require changes to the workflow? it wouldnt be called weekly but dream instead since it reminds of the anthropics dreaming. good idea or not?
18.
p10: but how can i do the manual terminal commands if they are worktrees? or do the two commands i usually run in prs work then too fine? will the p10 automatically use worktrees when i tell claude to start working on a feat x?
19.
wait so i can only test/build one pr/session/repo at a given time?
20.
[a request to measure a local stack's memory and CPU use on this Mac left out]
21.
p16 why no bypass mode?
22.
p25 i have quite many [a downstream project] sessions open cant i just do that still because theres usually always open some sessions on [it]?
23.
commit the weekly changes and rename weekly to dream and anything else i should do about the dream workflow? does it cover "redunant" things that i have in my dotclaude that have already been incorporated to the claude default harness by anthropic?
2026-10-16, commit 0cf6598, model claude-opus-5-5, effort xhigh.
1. Quoted under 27b102b.
2. (answer to which message was meant)
[an ask about a downstream project left out]
3. (a message from another Claude session, not typed by the user)
[a note about a downstream project's memory left out]
4.
dotclaude retry atleast once always if you cant open a file from slack connector or how many tries did you do?
5.
lets debug why slack connector cant open file
6.
should we add a rule about ext and builtinbrowsers to dotclaude?
7. (answer to whether to add the merged rule and push)
go
2026-10-16, commit 5239b0d, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit d6a6580, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit 6d420c0, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit 5542c3b, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit 9d48081, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit dddd06c, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit fac48a0, model claude-opus-5-5, effort xhigh.
1.
regenerate p18 from the weekly report
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit 57ae11d, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit f49c2be, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit 022aead, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit 8136ad9, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit b744ac0, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit ee21883, model claude-opus-5-5, effort xhigh.
2. Quoted under bd95b64.
3. (answer to whether the scripts in the terminal were done) Quoted under bd95b64.
2026-10-16, commit bd95b64, model claude-opus-5-5, effort xhigh.
2.
go, run p18 and commit and push dotclaude
3. (answer to whether the scripts in the terminal were done)
done running scripts, run p18 then commit and push everything pending in dotclaude, one commit per proposal
2026-10-15, commit 8974a1e, model claude-opus-5-5, effort xhigh.
1.
regenerate p9 from the weekly report
2. Quoted under 168c498.
2026-10-15, commit 2363a7e, model claude-opus-5-5, effort xhigh.
2. Quoted under 168c498.
2026-10-15, commit b2d761e, model claude-opus-5-5, effort xhigh.
2. Quoted under 168c498.
2026-10-15, commit e03c79c, model claude-opus-5-5, effort xhigh.
2. Quoted under 168c498.
2026-10-15, commit 125e1b5, model claude-opus-5-5, effort xhigh.
2. Quoted under 168c498.
2026-10-15, commit dfc4754, model claude-opus-5-5, effort xhigh.
2. Quoted under 168c498.
2026-10-15, commit 9c39a2c, model claude-opus-5-5, effort xhigh.
2. Quoted under 168c498.
2026-10-15, commit 168c498, model claude-opus-5-5, effort xhigh.
2.
go, run p9 and commit and push dotclaude
2026-10-15, commit 6f9607e, model claude-opus-5-5, effort xhigh.
16. Quoted under 696018f.
2026-10-15, commit 696018f, model claude-opus-5-5, effort xhigh.
16.
force push with lease to dotclaude. dotclaude add commands that i can run or copy with the proposals in the response. also the order is wrong, youre now explaining those so that for example p2 should be run after p6, p7 etc. change numbering in the future maybe so that p1 is the first that should be done. its good that you mention severities before going through the list of what i must do next
2026-10-15, commit e576a66, model claude-opus-5-5, effort xhigh.
5.
why 5 commands and not 2 like other two prs had? dotclaude fix. [an ask about a downstream project's pull request template left out] dotclaude and also this pr move the what has changed sentence to the top, then closing issues, then the prompts etc and then verify. [an ask to move that template's checklist into the project's agent docs left out]
2026-10-15, commit 6650f14, model claude-opus-5-5, effort xhigh.
15. Quoted under 24b09c5.
2026-10-15, commit 24b09c5, model claude-opus-5-5, effort xhigh.
15.
fix the two things. anonymize and dotclaude so that it doesnt happen again. list the whole weekly report here always im not going to open that copy anywhere else
2026-10-14, commit ae4d0d9, model claude-opus-5-5, effort xhigh.
12.
do you think this particular workflow should be js then instead of shell script? why not ts? should my hooks be js as well?
13.
so if i were to run the weekly now would it stop at 95% 5h? do i have to separately tell it to not continue until weekly usage is 95%?
14.
/weekly
2026-10-14, commit c397f0f, model claude-opus-5-5, effort xhigh.
10.
remove [a person] and [another person] from the workflow
11.
if i fire the weekly now (also rename it as weekly), will it put my 5h usage to 95% ish
2026-10-14, commit cce004a, model claude-opus-5-5, effort xhigh.
7.
github plugin removed from claude.ai. how much will opus 5.5. extrahigh premium team seat usage cost in the weekly review
8.
what do you mean by proposals? also my limits are [pasted: the usage limits panel from the app] for some reason on sat and not friday.
9.
right and can i somehow modify the workflow so that it collects data on how much usually x amount of proposals take usage and with time the workflow will always look at my usage and use the estimated amount of proposals so that usage fill doesnt block and stop the workflow?
2026-10-14, commit 4f573a5, model claude-opus-5-5, effort xhigh.
6. Quoted under 67172c0.
2026-10-14, commit 67172c0, model claude-opus-5-5, effort xhigh.
4.
do i need the terminal claude to anything? are you srue github plugins just fails to connect 36 of 41 sessions because i have vpn on? how to fix the figm aplugin thingy?
5.
can i just uninstall claude code terminal?
6.
switch the doctor step to the desktop build. also i uninstalled before doing the figma commands
2026-10-14, commit 03071ff, model claude-opus-5-5, effort xhigh.
6.
[pasted: a review of why the screenshot upload stalled, asking for the upload through the built-in browser, a memory index fix and a fix to this rule]
2026-10-14, commit 8f63649, model claude-opus-5-5, effort xhigh.
4.
[an ask about a downstream project left out]
5.
dotclaude always upload the screenshots and recordings to the pr if they exist and are relevant in the pr
2026-10-14, commit c97a0a4, model claude-opus-5-5, effort xhigh.
1.
will / doctor analyze deprecations based on my claude history or based on anthropic patch notes
2.
do you think i could be able to build a thing (help me find a better word for this? skill? hook? workflow?) that i can run inside claude at the end of week maybe due to resets over weekend that spends a lot of tokens in analyzing anything it has access to in my computer to make my dotclaude better and tell me what to do differently nex week? maybe fetch [a person's] skills repo and check if anything would be useful to me? maybe fetch [another person's] sklils etc and do the same? maybe use /doctor and the prompt audit as well? maybe fetch anthorpic news / pathc notes / changelog to spot something i should do differently?
3.
can you create a workflow like this for me? how are workflows different than skills? i dont wanna schedule it i just wanna run it myself in case i have tokens left at the end of the week
2026-10-13, commit bf8105e, model claude-fable-5-1, effort xhigh.
15.
dotclaude never reply to anyone on behalf of me on github
2026-10-13, commit 27b102b, model claude-fable-5-1, effort xhigh.
1.
[an ask about a downstream project left out]
2.
[an ask about a downstream project left out]
3.
dotclaude never leave stack up
2026-10-13, commit e2a5a2a, model claude-fable-5-1, effort xhigh.
13.
dotclaude so i want the other non checked box to be deleted isntead of jsut left unchecked
2026-10-13, commit 33f590f, model claude-fable-5-1, effort xhigh.
12.
dotclaude you didnt put url localhost etc in the pr description
2026-10-13, commit 95b332c, model claude-fable-5-1, effort xhigh.
11.
dotclaude put the images and videos under the step number not at the same line
2026-10-13, commit 4b45b7b, model claude-fable-5-1, effort xhigh.
10.
dotclaude pull out the screenshots and videos from the detail element in pr description and put them under the steps where they are used. preferably at the start of the step since were not too sure when the screenshot happens exactly and what to look for. can you slow down the videos since they are playwright automation or not possible?
2026-10-13, commit 644646a, model claude-fable-5-1, effort xhigh.
8.
[an ask about a downstream project left out] dotclaude dont open two prs in one session unless i explicitly state so
2026-10-12, commit 36b2e16, model claude-fable-5-1, effort xhigh.
11.
dotclaude. a step’s screenshot is named at the start of the step and comes before the ok that closes what it shows, pr 110 put both after the ok and the save. confirm first before pushing.
also dotclaude in the pr instruction missed section give me a github md version of the commands that you gave me as instructions like in this case from 1 to 5 steps. confirm first
12.
ok now give the markdown of them steps
13.
push
2026-10-12, commit d5727d3, model claude-fable-5-1, effort xhigh.
7.
dotclaude give me the steps you would have given me to test things. in the end of my "review pr #" prompt tell me what the pr instructions missed etc like the seeding of already tested workflows instead if me clicking through things. when a step has a video or image attached to it in pr tell me to look at this point at the pr. also the () approach of not doing cd .. seems good. shoudl we also do the way its done in the pr? also is it good practice to chain other commands in the instructions you just gave me? thoughts before commit?
8.
not good enough. you missed shell commands in the pr 110 steps. dotclaude fix. confirm first
9.
retry pr 110 review steps
10.
excellent. push the dotclaude edits that made your response so good
2026-10-12, commit 719067c, model claude-fable-5-1, effort xhigh.
6.
[a mirror] now has commit message such as "[a content type] with a few children" which gives information regarding a private customer project and gives unnecessary information regarding the actual commit to [a mirror]/dotclaude. dotclaude: anonymize in future somehow or remove unnecessary bits
2026-10-12, commit 986a428, model claude-fable-5-1, effort xhigh.
5.
no. dotclaude dont put copypastable things to pr reviews. just tell me shortly what should be corrected
2026-10-12, commit a9604bf, model claude-fable-5-1, effort xhigh.
4.
[pasted: the previous script's terminal output]
didnt seem to push anything to [a mirror]. also dotclaude never give me python scripts when doing thing on shell. give me shell script.
also give me the corrected pr description instructions on how to test the feat. and then give me instructions where the pr should have mentioned somethign else. put this also in dotclaude
2026-10-12, commit b45b966, model claude-fable-5-1, effort xhigh.
1.
review pr 110 here only give manual testing instructions
2.
you literally said in pr 12 review "review pr 12 here only, give manual testing instructions, and leave the checkout on its branch”" is the optimal way to say this.
how should i have said it so that you would have understood to run also verify, e2e and literally everything you can before i manually test this
3.
has the author not made a manual testing seed that contains the "[a content type] with a few children" because putting that would not reflect the staging backend of [the backend vendor] and would thus create tests that would break in gcp when tested against staging backend e2e? why not create a different seed for e2e tests and a different one for manual local testing?
also dotclaude: dont put your arbitrary instructions on how to test here but rather just corrections to the pr instructions if any. your commands are nice tho keep them
2026-10-10, commit 92bccf0, model claude-fable-5-1, effort xhigh.
2.
well then change the rules of the checkout local identity. my name is fine in history. others ofc not
2026-10-10, commit 9d7fdc6, model claude-fable-5-1, effort xhigh.
1.
add local project memory to sync also to [a directory in a second repository]? i was thinking on local project memory so that public dotclaude users wont see the [second repository] path? confirm before doing anything
Script mirror + memory (AskUserQuestion selection) Yes, commit and push to main (AskUserQuestion selection)
2026-10-10, commit aa3cc21, model claude-fable-5-1, effort xhigh.
4.
› what in gods name is this character. dotclaude dont output chars like that. also why name them pr12 when they could have been more shorter like 123 or abc? dotclaude fix also "Paste pr12 c’s address into a new tab". you never told me to copy the address. fix in dotclaude
confirm all with me before doing anything
5.
yeah sure commit and push
2026-10-10, commit 8a2f66e, model claude-fable-5-1, effort xhigh.
1.
review pr 12 and give manual testing instructions here
2.
naming it a comes before setting the company as company1-sub. this "wrong order" happens other times in the instructions you gave me. fix in dotclaude so that this never happens again.
On A, pick any other content, then click C’s name. what? what do you mean pick any other content? also cancel is back in the ui after clicking C. also when clicking the contents breadcrump i.e. /contents theres no back button. "both ask too"? what? > Remove the picked content again . what? ohh by pick you mean Add item.
ok picking a content is just wrong if you mean Add item, which is adding content, although in some sense yes youre picking content logically but in ui its called adding. obviously this finnish thing is not working so research and figure out if we should make a change in dotclaude like click "Add item" button name instead of calling things with their logical names. dont commit anything without confirming with me first
3.
yeah commit and push dotclaude
2026-10-09, commit 59970a4, model claude-fable-5-1, effort xhigh.
7.
global ruling: if theres two ways to do something dont say 1, 2 etc. say a or b right? thoughts?
also retry. cant you open the browser yourself inside claude chrome?
2026-10-09, commit a6451f2, model claude-fable-5-1, effort xhigh.
6.
are you able to take screenshots or screen recordings of when for example the playwright runs to the pull request description?
7.
can you put a rule in my dotclaude to put any screenshots and recordings to the pull request in details element if the pr has playwright and pr has the purple thingy screenshots section? thoughts before committing?
8.
yeah change visual. ive logged in the calude chrome on github so use that. thoughts?
2026-10-08, commit 5f8d268, model claude-fable-5-1, effort xhigh.
4.
global ruling: if classifier refuses something i can fix with giving more permissions tell me the shell script that i can fix it with
5. (confirming that both handed-over scripts were run)
ran
2026-10-08, commit f3b322b, model claude-fable-5-1, effort xhigh.
2.
ok lets continue working on the [a feature of a downstream project]. also do you need more settings perms to do gh issue edits?
3.
i redid that in zsh so you should have more perms. go ahead
2026-10-08, commit f1fa780, model claude-fable-5-1, effort xhigh.
14.
global ruling: put the merge prompt also to the pr when i say merge
2026-10-08, commit 280fcd7, model claude-fable-5-1, effort xhigh.
1.
[a question about a downstream project's backlog and next task]
global ruling: assign me to the issues a draft pr or pr closes if i have perms to assign myself to that issue
2026-10-08, commit 1dc8ca2, model claude-fable-5-1, effort xhigh.
7.
also you always start the stack not me. also so hotspot is now ok from backend code? what about other logs endpoints?
8.
yeah add the permission rule but you should probably not run it in background since you always complain about permissions not being able to stop it. or is that permission thing something i can fix as well?
9.
convention is stack stays up? huh? but when i review prs or do my own pr i think its good practice to stop the stack for example if i have to install something and cant be hot reloaded? thoughts? also give me sed command etc to edit the settings.json
10.
done, go with your recommendations on all three
2026-10-07, commit 8d49c37, model claude-fable-5-1, effort xhigh.
1.
whats that one chatgpt session where this mathematician speaks very technical with chatgpt on maths and it sets the skill level the same way? can i somehow elevate my claude sessions like that since im a cs masters graduate?
2.
any proof there was a register.md or similar with tao? also look at my claude code history and analyze if i would actually benefit from a register.md like that?
3.
yeah sure sync it
2026-10-07, commit 867d6f2, model claude-fable-5-1, effort xhigh.
6.
global memory: dont put e2e tests then never after manual testing in prs unless said otherwise
2026-10-07, commit f4cb122, model claude-fable-5-1, effort xhigh.
5.
global memory: dont tell me step 9 again pr review scenarios or whenever a pr checkout was used as the step 2 anyway rewrites that each time i review pr
also why is the e2e spec after testing manually?
2026-10-07, commit 1e889d4, model claude-fable-5-1, effort xhigh.
4.
global ruling: no if conditions on like that on submodule. just update submodule also project memory: clear the symfony cache always no need to do conditional guesswork
also none of those commands starts the whole thing like [cli] start or [cli] develop. not too professional
2026-10-07, commit 62b6d96, model claude-fable-5-1, effort xhigh.
1.
review pr 104 and tell me how to test locally
2.
dont i have to update backend module if the last pr i reviewed or created had a different version in it?
3.
global ruling: when switching branches always check and update the backend to the version its in the checked out branch
2026-10-07, commit a861b43, model claude-fable-5-1, effort xhigh.
4.
the cd playwright thingy is again without the cd back. also no need for numbering nor explanatiosn nor check by hand section
2026-10-07, commit 5af1440, model claude-fable-5-1, effort xhigh.
2.
global ruling: when doing cd operations remember to cd back to the place where it cd from
also why would i have to seed rows back afterwards? im gonna reseed after each new pr anyway.
also project memory maybe: you should probably not start [the project's cli] things in background since you dont have permissions to stop them yourself afterwards. thoughts?
2026-10-06, commit 2fbeea0, model claude-fable-5-1, effort xhigh.
1.
i would like claude to remove stale branches i.e. that dont have pr or draft open in them when sending a new prompt or session? thoughts
2.
does my current workflow already remove remotely deleted branches from local?
3.
ok lets do the remove safely branches that are deleted from remote already in the beginning of the session
2026-10-06, commit 0dbe828, model claude-fable-5-1, effort xhigh.
2.
add gh repo create for bash in my claude code settings
2026-10-03, commit d246b5c, model claude-fable-5-1, effort xhigh.
2.
put the places where to check to project memory so that i i can just say lets start doing feat x without having to list all the resources youre going to look at always before doing that. youll probably also fetch figma only when it has some edits in there and not all the time. also what is one draft pr at a time? i probably meant like only one draft per conversation/feat but this is a different feat than those two
2026-10-03, commit 0e6c226, model claude-fable-5-1, effort xhigh.
7.
global claude rule you can use semicolons
2026-10-03, commit b05ac24, model claude-fable-5-1, effort xhigh.
3.
global ruling: you can still use " quotes in the prompt feedback and shift characters. just dont use them at the rewritten part. you can still do like ReWritten: "blahblah" but not Rewritten: "Blj#445K". get it?
also i dont see the new 10 findings i only see the 8 in the old response
2026-10-03, commit 4ae697a, model claude-fable-5-1, effort xhigh.
2.
pull the branch and run the same review here locally results. was probably force pushed. also ""Review PR 94 against the local stack: run verify and the content e2e specs, click through its How to verify steps, and report findings here rather than on GitHub."" how to say this in lowercase?
global ruling: always give prompt feedback rewritten as lowecase and without using shift operand if possible and as little typing as is to make the prompt actually good but not overly long
2026-09-30, commit d7dd1ad, model claude-fable-5-1[1m], effort xhigh.
4.
global ruling: instead of a details element link my dotclaude there in copyable code block. i think blue clickable links are antipattern and might lead to somewhere other than the visual url. can someone put malicious invisible characters to a copyable code block on github markdown?
2026-09-29, commit 9e0b37c, model claude-fable-5-1[1m], effort xhigh.
2.
global ruling: why did you not put the tool start --clean in copyable triple backticks? also recreate mock database did not contain a command. why? fix in global ruling so that this problem never repeats again
2026-09-29, commit a2762f8, model claude-fable-5-1, effort xhigh.
5.
global ruling: put co-authored of any reviewers/commenters of the pr when merging the pr too. you can for example find the git email and name from git log grep? if you cant find the exact match dont put it into the coauthored by. thoughts before committing to global ruling?
6.
yeah the reviewed by approach sounds good. apply to the global ruling.
next thing to global ruling: can i somehow apply the projects own internal md like for example where the discovery lives (not in git) to the pr description before or after the claude.md line, whichever is the order that claude code reads them. thoughts before committing to global ruling?
2026-09-28, commit 15c58c6, model claude-fable-5-1, effort xhigh.
2.
nooo wayyy too verbose. like one sentence per each item to be explained
2026-09-28, commit 3eedb3f, model claude-fable-5-1, effort xhigh.
1.
explain in readme what everything/every file does. for example it was unclear what the sync.sh does just based on the name. or the git fetch prune hook
2026-09-27, commit 7786fa1, model claude-fable-5-1, effort xhigh.
4.
named in passing? also now the commands are in one block in pr and will probably fail because no ampersand or no new codeblock. always figure out which one to use.
2026-09-27, commit 2d8439b, model claude-fable-5-1, effort xhigh.
3.
global ruling: when writing terminal commands to the mds like the docker compose one in the pr, put it int triple backticks etc. so that copying and reading is easier
2026-09-26, commit 1423306, model claude-fable-5-1, effort xhigh.
2.
global ruling: use git switch whenever you can get away with it rather than opting for checkout because checkout can do more than just switch branches
2026-09-25, commit 7460cd2, model claude-fable-5-1, effort xhigh.
1.
global ruling: when squashing pull requests remember to add co authored by the proper human accounts that were part of that pr commits
2026-09-25, commit c41d5c6, model claude-fable-5-1, effort xhigh.
8.
you forgot to put prompt to the description even though we didnt push commits
9.
yes but why did you miss it? did i not state it in global ruling good enough?
10.
sync it
2026-09-24, commit 33c5067, model claude-fable-5-1, effort xhigh.
2.
remember: dont put abbreviations details element to commit messages nor to pr descriptions if there are none
2026-09-24, commit 1cba951, model claude-fable-5-1, effort xhigh.
1.
can you write in my dotclaude repo so it reads as mine e.g. a public mirror my ~.claude etc. also do we really need that copyright lambdakilo 2026 thingy? do we need those spdx identifier phrase either? also can you with gh put the correct topics tagged in the repo?
2026-09-23, commit 49f1c57, model claude-fable-5-1, effort xhigh.
1.
can you put into my dotclaude that the model should at the end of reply tell me how my prompt could have been better
2026-09-23, commit 975b506, model claude-fable-5-1, effort xhigh.
4.
ok then commit that and push settings.json change
2026-09-23, commit d8ce80a, model claude-fable-5-1, effort xhigh.
3.
the commit message always abbreviates llm. dont abbreviate that or just remove the word llm. the pr prompt text doesnt even abbreviate pr (and thats good)
2026-09-23, commit 6df3bb5, model claude-fable-5-1, effort xhigh.
2.
remember begin: if the pr template has the added tests? just put the check on the one thats us for example if youve written tests then thats it if not write one sentence why not. in the what has changed write one sentence on what has changed. write the related ticekts and documents. do the same thing always with the is documentation up to date thingy. also write the how to verify (manually) .
[a second paragraph asking for a pull request to a template repository, not about this file]
2026-09-22, commit 0d8a7d9, model claude-fable-5-1, effort xhigh.
4.
did you not read the global ruling claude.md? the part of not putting prompts first in commit messages? or what happened here? help me debug, but dont force push anything rn
5.
whats a commit rule
6.
yeah put a line in the commit rules
2026-09-21, commit ad33df5, model claude-fable-5-1, effort xhigh.
3.
should the CLAUDE.md file be called something else than that so that my sessions started inside that local clone wont load both my claude.md from .claude and the dotclaude as well? any other files that need to be named otherwise? im not sure about the copyright email thingy in readme. plus license is h1 although a haeder 1 should be my .claude folder etc. ah you must have misunderstood my [address] email thingy there. i meant it to be used in git config email and nowhere else. does github still show my commits verified although i cant add [pasted: an email address] as my github email but i can set it as my git email? also you censor well copypastes in pull requests but in commit messages they still are not removed. fix in global ruling. take a look at [a downstream project] latest merge commit we made. also censor that name here. also maybe limit commit message widths to 80 or something more standard? they break pretty bad in github interface. also in commit message maybe put my prompts only after the commit message itself but in pr its ok to put them first. make these global ruling
2026-09-19, commit d717a6c, model claude-fable-5-1, effort xhigh.
2. Quoted under df2d3d7.
2026-09-19, commit df2d3d7, model claude-fable-5-1, effort xhigh.
2.
actually yeah put the prompts in all of my commits in the future. git commit messages also should probably say that this is the llm prompt and which model and effort was used. remove the readme part except the license part. just put my disposable email [pasted: an email address] since github is a pain in the ahh with simplelogin. rename the repo dotclaude that makes more sense. are you sure the shell scripts wont drift with time? im planning to just tell claude on my arch desktop (rather than this m5 air) to fetch my dotclaude once and then itll sync itself with gh and git etc. tell me if my intuition is wrong
2026-09-19, commit a6bd820, model claude-fable-5-1, effort xhigh.
1.
how to set global ruling (is that what its called in short? the ~.claude/claude.md) that gitch fetch --prune should be run from time to time when in git repos? at each prompt? also create a repo on my lambdakilo account that contains stuff like my claude.md and anythign else that might be useful for learning about my workflow for other people. repo name can be for example .claude. dont create readme. i wanna use either agplv3 or later (so actually create the appropriate or later message in readme after all lol for agplv3 if we choose agplv3) or ccbysa4.0. use both or either or neither whichever is applicable for our work. put my prompts in git commit messages. always update the .claude github repo whenever theres changes in local .claude etc my claude settings useful for others. never put my name or any company names or secrets or anything disclosing on what im working with in the .claude repo. also now that were on the subject of my claude.md: global ruling: never write pull request descriptions, let me write them myself. also always put a details element list of abbreviations in the prompt details element i might have used in the prompts so that people can open and check those if i use some abbreviations people dont know about
- 5h: 5-hour
- 5h: five hours
- agplv3: GNU Affero General Public License, version 3
- ai: artificial intelligence
- arch: Arch Linux
- btw: by the way
- ccbysa4.0: Creative Commons Attribution-ShareAlike 4.0 International
- cd: change directory, the shell command
- char: character
- co2: carbon dioxide
- cs: computer science
- cuz: because
- dev: developer
- dir: directory
- e.g.: for example
- e2e: end to end
- esim.: esimerkiksi, for example
- etc: et cetera
- eu: European Union
- evals: evaluations
- exfil: exfiltrate
- ext: external
- ext: external, a file posted by another organisation in a shared channel
- feat: feature
- ff-only: fast-forward only
- fri: Friday
- gcp: Google Cloud Platform
- gh: GitHub CLI
- gh: GitHub command line interface
- h1: HTML heading level 1
- html: HyperText Markup Language
- i.e.: id est, that is
- i.e.: that is
- ip: intellectual property
- js: JavaScript
- kwh: kilowatt-hours
- LLM: large language model
- lol: laughing out loud
- m5 air: MacBook Air with the M5 chip
- md: Markdown
- md: Markdown file
- mds: Markdown files
- mon: Monday
- ofc: of course
- oy: osakeyhtiö, the Finnish limited company form
- p18: proposal 18 of the weekly review of 2026-10-14
- p9: proposal 9 of the dream report of 2026-10-17
- p9: proposal 9 of the weekly review of 2026-10-14
- pdf: Portable Document Format
- perms: permissions
- pls: please
- PR: pull request
- prs: pull requests
- qr: quick response code
- repo: repository
- rest: REST, an HTTP API style
- rn: right now
- sat: Saturday
- sh: shell script
- sod: start of day
- spdx: Software Package Data Exchange
- ss: screenshot
- tmp: the temporary directory, /tmp
- ts: TypeScript
- ui: user interface
- uri: uniform resource identifier
- url: uniform resource locator
- usd: US dollars
- vpn: virtual private network
- wth: what the hell
- XDG: X Desktop Group, now freedesktop.org