kunalsuri · kunalsuri · Jul 2, 2026 · Jul 2, 2026 · Jul 2, 2026
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
@@ -33,3 +33,20 @@ jobs:
         run: python -m py_compile install.py lib/__init__.py lib/drift.py lib/installer.py lib/intake.py lib/maturity.py lib/orient.py lib/util.py lib/verify.py
       - name: Run smoke tests (both installers)
         run: node test/run-tests.mjs
+
+  # Coverage floor for the Node installer (c8/V8; the Python installer is
+  # exercised functionally by the matrix job above but not instrumented).
+  # Floors sit a few points under the measured numbers so environment noise
+  # doesn't flake the gate; raise them as coverage improves.
+  coverage:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+      - name: Measure coverage and enforce the floor
+        run: >
+          npx --yes c8 --include "lib/**/*.mjs" --include install.mjs
+          --check-coverage --lines 83 --branches 73 --functions 90 --statements 83
+          --reporter=text node test/run-tests.mjs
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -44,6 +44,34 @@ Format: [Keep a Changelog](https://keepachangelog.com/) · Versioning: [SemVer](
     (the `github/` → `.github/` mapping already covered the Copilot assets).
   - No new rules files were added for Antigravity: it already reads the tool-agnostic
     `AGENTS.md` at the repo root natively.
+- **Coverage-driven test-suite hardening.** The smoke suite grows from 246 to 327
+  checks (Node installer coverage: 78.6 % → 85.8 % lines, 66.6 % → 76.4 % branches):
+  - table-driven `orient` detector tests across every supported stack (Gradle both
+    variants, Go, Rust, Ruby, PHP, CMake, the bare-Makefile fallback and its
+    suppression, pip, yarn/bun/Pipenv lockfiles, Turborepo, `package.json` without
+    a `build` script, malformed `package.json`, empty repo, multi-stack note) plus
+    `--name`/`--description`/`--build`/`--test`/`--upstream` overrides and README
+    description-extraction edge cases (badge/HTML skipping, 160-char truncation,
+    fill-in fallback);
+  - per-ecosystem `indepth` dependency-parsing tests (pip, Poetry, Go, Cargo,
+    Bundler, Composer) and a real-git-repo fixture exercising the git-history
+    analyzer (commit/contributor/tag counts) in both runtimes;
+  - unit tests for `classifyAction` pinning the full re-run matrix — including the
+    kit-upgrade `"update"` path that end-to-end tests cannot reach — and for the
+    intake wizard's `detectBranch` (normal / detached HEAD / no `.git`; exported
+    from `lib/intake.mjs` for testing);
+  - `verify` edge cases (no knowledge docs → clear error; duplicate basenames →
+    `2 matches` note) and CLI surface checks (usage text, unknown option);
+  - a **Node ↔ Python parity test**: both installers must write byte-identical
+    orient profiles (modulo timestamp) for the same fixture, so cross-runtime
+    drift now fails the suite instead of shipping;
+  - `npm run coverage` (c8, fetched via `npx`) and a CI `coverage` job in
+    `test.yml` enforcing a floor (83 % lines / 73 % branches).
+
+### Fixed
+- Python `orient` recorded fork evidence with an ASCII `->` where the Node
+  installer writes `→`; the two now emit identical profiles (caught by the new
+  parity test).
 
 ## [0.1.2] — 2026-06-30
 

diff --git a/ai/analysis/audit-reports/DRIFT_REPORT.md b/ai/analysis/audit-reports/DRIFT_REPORT.md
@@ -11,5 +11,3 @@
 | stale | 0 | `[verified]` row whose code changed since the verified commit |
 
 No drift detected. The map matches the tree.
-
-> Stale check: stale check is opt-in — re-run with --git to compare against the last verified commit (local, read-only git).
diff --git a/ai/guide/MODULE_MAP.md b/ai/guide/MODULE_MAP.md
@@ -4,8 +4,9 @@
 > **Index only.** Find the area here, then open the entry file directly. Don't crawl
 > the tree. The directory list can be regenerated; **Responsibility** and **Stability**
 > are judgement and must be audited by a human.
-> Last verified: 2026-06-25 @ commit ba964e9 (drift baseline re-anchored 2026-07-02 to the
-> squash-merge commit on main — the per-row human audit dates below remain the authoritative signatures)
+> Last verified: 2026-06-25 @ commit 1fd11db (drift baseline re-anchored 2026-07-02: the
+> human-merged PRs #5–#7 changed `/`, `templates/`, and `docs/` after the previous baseline —
+> the per-row human audit dates below remain the authoritative signatures)
 
 > Status: drafted by an `[inferred]` cold-start pass on this repo (the kit dogfooding
 > itself), then audited by a human on 2026-06-25 who set each row's Stability to `ours`

diff --git a/lib/intake.mjs b/lib/intake.mjs
@@ -27,7 +27,8 @@ const FAMILIARITY = [
   "Not sure",
 ];
 
-async function detectBranch(targetAbs) {
+/** @internal exported for testing. */
+export async function detectBranch(targetAbs) {
   // .git/HEAD is "ref: refs/heads/<branch>" on a normal checkout. Reading it is
   // file inspection, not a git command — we never shell out.
   const head = await readText(path.join(targetAbs, ".git", "HEAD"));

diff --git a/lib/orient.py b/lib/orient.py
@@ -55,7 +55,7 @@ def detect_fork(target, flags):
             url = m.group(1)
             gh = re.search(r'[:/]([^/:]+/[^/]+?)(?:\.git)?$', url)
             return {"isFork": True, "upstream": gh.group(1) if gh else url,
-                    "evidence": f'git remote "upstream" -> {url}'}
+                    "evidence": f'git remote "upstream" → {url}'}
     return {"isFork": False, "upstream": None, "evidence": 'no remote named "upstream"'}
 
 

diff --git a/package.json b/package.json
@@ -24,7 +24,8 @@
     "setup": "node install.mjs install",
     "remove": "node install.mjs uninstall",
     "test": "node test/run-tests.mjs",
-    "deep-test": "node test/run-deep-test.mjs"
+    "deep-test": "node test/run-deep-test.mjs",
+    "coverage": "npx --yes c8 --include \"lib/**/*.mjs\" --include install.mjs --reporter=text node test/run-tests.mjs"
   },
   "repository": {
     "type": "git",

diff --git a/test/README.md b/test/README.md
@@ -9,6 +9,7 @@ This directory contains the testing suites for the `ai-fication-kit` repository.
 |---|---|---|---|
 | **`npm test`**<br>(`node test/run-tests.mjs`) | Installer Functionality | Runs cross-runtime smoke tests for the Node and Python installers against throwaway mock repositories (e.g. testing orient, install, shazam, and uninstall). | Run during active development of the installer code or stack detection modules. |
 | **`npm run deep-test`**<br>(`node test/run-deep-test.mjs`) | Full Repo Health & Standards | Performs a complete audit of the repository, including running the installer smoke tests, document claim verification, structural drift detection, license header checking, and placeholder leakage detection. | Run before submitting a PR, tagging a release, or declaring a task complete. |
+| **`npm run coverage`**<br>(c8 over `run-tests.mjs`) | Node Installer Coverage | Runs the smoke tests under [c8](https://github.com/bcoe/c8) (V8 coverage, fetched via `npx`; the child `install.mjs` processes are instrumented too) and prints per-module line/branch coverage for `install.mjs` and `lib/*.mjs`. CI enforces a floor in the `coverage` job of `test.yml`. | Run when adding tests or touching `lib/` to see what the suite actually exercises. |
 
 ## Details of Deep Test Verification
Original file line number	Diff line number	Diff line change
Expand Up		@@ -11,5 +11,3 @@
		\| stale \| 0 \| `[verified]` row whose code changed since the verified commit \|

		No drift detected. The map matches the tree.

		> Stale check: stale check is opt-in — re-run with --git to compare against the last verified commit (local, read-only git).