Three-layer skill set built on the methodology of @darkseoking — an SEO practitioner and Threads creator with 10+ years of hands-on experience, 17k+ followers grown through algorithm-informed content strategy. He spent real money on controlled experiments and reverse-engineered Meta/Google algorithm patents to understand how content distribution actually works. Every insight in these skills is backed by patent filings or tested data — no speculation, no "industry consensus" parroting.
Knowledge → darkseoking-mindset 8 mental models + 8 reference docs
Operations → darkseoking-post-optimizer Pre-publish checklist (PASS/WARN/FAIL)
Prediction → darkseoking-post-predictor Engagement ceiling estimation
Triggers: SEO algorithm questions, content strategy, Threads growth, GEO optimization, vendor evaluation
The knowledge layer. Eight core mental models (reverse-engineer from patents, pay to test, question consensus, cross-platform logic, boundary conditions, minimize cost, toolify knowledge, account positioning) plus eight deep-dive reference docs covering algorithm analysis, content strategy, Threads growth, GEO/AI citation, vendor evaluation, AI tooling, experiment design, and operational philosophy.
Triggers: Writing/reviewing Threads posts, post-viral strategy, posting timing
The operations layer. Runs an 8-item pre-publish checklist — each item rated PASS/WARN/FAIL with specific reasoning:
- First Line (info density, no clickbait)
- Semantic Precision (domain-specific terms)
- Topic Distance (diversity penalty avoidance)
- External Links (authority + relevance)
- Thread Structure (standalone main + value-add threads)
- Timing (self-competition avoidance)
- Post-Viral Strategy (only after viral post)
- Topic Suggestion (only when asked "what to write")
A marketing consultant's account focuses on SEO and AI tools. Their last post (12 hours ago) about "5 AI writing tools" got 800 likes — well above their usual 200. They now want to post this:
Draft:
AI is changing everything about how we do SEO. Here are 3 more tools that will blow your mind and transform your content workflow forever. Trust me, these are game-changers! 🚀🔥
Link: bestaitools.xyz/my-affiliate-list
Checklist result:
| # | Check | Rating | Reason |
|---|---|---|---|
| 1 | First Line | FAIL | "AI is changing everything" is a textbook AI-typical opener with zero info density. "Blow your mind" and "game-changers" are clickbait filler. |
| 2 | Semantic Precision | FAIL | Dominated by generic terms ("AI", "SEO", "content workflow"). No domain-specific nouns — which tools? what kind of SEO? Nothing makes the target audience think "this is for me." |
| 3 | Topic Distance | FAIL | Same topic (AI tools) + same angle (listicle) as the post 12 hours ago. Diversity mechanism (US9336553B2) will penalize same-source density. |
| 4 | External Links | FAIL | Unknown domain (bestaitools.xyz), likely affiliate link. Low authority + content relevance unclear. |
| 5 | Thread Structure | WARN | "3 tools" has enough content for threads, but is packed into a single post. Splitting into 1 tool per thread reply would add independent information value per reply. |
| 6 | Timing | FAIL | Last post was 12 hours ago and is still in active diffusion (800 likes = well above baseline). Posting now competes with yourself (US8402094B2). |
| 7 | Post-Viral Strategy | FAIL | Last post went semi-viral. Should be engaging with comments on that post, not rushing a follow-up on the same topic. |
Result: 1 WARN, 6 FAIL — do not publish.
Optimizer recommendations:
- Wait 24-48h. Let the 800-like post finish its diffusion cycle. Use that time to reply to every comment on it instead.
- Change topic. If the account covers SEO + AI tools, the next post should lean into SEO (different topic, same niche) rather than another AI tools listicle.
- Kill the affiliate link. Either remove it entirely (safest) or replace with a high-authority source that directly supports the content.
- Rewrite the opener. Lead with a specific, counterintuitive finding — not "AI is changing everything."
Rewritten (after waiting 48h):
I used to mass-produce SEO articles with AI — 50 per week. Rankings went up for 2 months, then every single one dropped out of the top 100. Here's what the data showed was killing them, and the 3-step fix that brought 80% back.
[Thread 1] The first signal was... [specific data] [Thread 2] The fix that worked... [Thread 3] What I'd do differently now...
Changed topic (AI content quality, not tools listicle), led with a counterintuitive first-hand result, removed affiliate link, added thread structure for depth.
Triggers: Predicting post performance, analyzing history, estimating engagement ceiling
The prediction layer. Estimates engagement ceiling for a draft, with or without personal post history data. With data (15+ posts): builds personal baseline, extracts patterns, predicts against them. Without data: uses darkseoking's 84-post benchmark as fallback for directional guidance.
darkseoking-all-posts.csv— 84 posts with engagement metrics (likes, reposts, comments, thread count, content), used as benchmark data across all three skills
| Patent | Mechanism |
|---|---|
| US10579688B2 | Semantic vector matching for content distribution |
| US9336553B2 | Diversity mechanism (penalizes same-source density) |
| US10558714B2 | Creator Embedding (account positioning) |
| US9378529B2 | Expected Engagement Value |
| US8402094B2 | Self-competition penalty for rapid posting |
| US10565267B1 | Rapid posting distribution penalty |
| US20140172877A1 | Priority distribution pool after pause |
| EP2977948A1 | Domain authority evaluation for links |
| US10268763B2 | Content relevance scoring for links |
New to Claude Code skills? See the ai-toolkit main README for general setup instructions.
- Copy the three skill directories (
darkseoking-mindset/,darkseoking-post-optimizer/,darkseoking-post-predictor/) into your project's.claude/skills/ - Place
darkseoking-all-posts.csvwhere the skills can reference it (e.g., alongside the skill directories, then adjust paths in SKILL.md) - Claude Code will auto-detect the skills based on their
descriptiontriggers — no manual activation needed
Each skill works independently but they complement each other:
- mindset gives you the "why" — algorithm mechanics and strategy principles
- optimizer gives you the "check" — systematic review before publishing
- predictor gives you the "what if" — performance estimation before committing