chore(andreessen): plugin-audit follow-ups — gemini sync + inner README#737
Conversation
Post-merge audit fixes for the andreessen productivity skill: - Add skills/andreessen/README.md (inner skill README). Lifts structure 91.3/EXCELLENT and quality 56.2; the README was the one genuine doc gap vs sibling skills. - Run the gemini cross-platform sync (codex ran at merge time; gemini was missed). Adds the andreessen symlink + index entry and reconciles a pre-existing stale claude-coach entry the generator surfaced. Audit verdict: PASS WITH WARNINGS. Structure 91.3 EXCELLENT, scripts 3/3 PASS, security PASS (0 critical/0 high), marketplace + ecosystem clean. The single warning is the quality scorer's title-case section/frontmatter schema that no Path-B productivity skill uses (andreessen 56.2 vs merged siblings reflect 44.6 / capture 46.4). https://claude.ai/code/session_01SF6MzfjHurZMt5JUFET9h3
… -> 65.7) Genuine, repo-consistent additions that lift the real quality gaps (not doc padding): - assets/forcing_question_worksheet.md — fillable 6-question interrogation - assets/blank_3x5_card.md — blank daily card template - assets/example_market_verdict.md — full worked market-first verdict - assets/example_pmf_check.md — worked before/after PMF check Both worked examples' tool invocations are verified against the actual scripts (MARKET-FIRST-DERISK at composite 6.36; BEFORE-PMF at composite 4.35). SKILL.md Assets section updated to reference all five. Quality scorer: 56.2 -> 65.7 (clears the 60 gate). examples 60->100%, assets 12->60%, practical_examples 40->80%. Audit now a clean PASS on quality alongside structure 91.3/EXCELLENT, scripts 3/3, security 0/0. https://claude.ai/code/session_01SF6MzfjHurZMt5JUFET9h3
- Bump headline counts: 329 -> 330 skills, ~448 -> ~451 tools, ~586 -> ~590 references, 49+ -> 50+ agents, 79+ -> 81+ commands across CLAUDE.md, README, docs/index.md, docs/getting-started.md, mkdocs.yml. - README productivity row 5 -> 6; add andreessen entry. - CLAUDE.md: v2.8.3 highlights section + version/status/date footer. - marketplace.json metadata + top-level description counts incremented. - mkdocs.yml nav: add andreessen skill page, cs-andreessen agent, /cs:andreessen + /cs:pmf-check commands. - Regenerate docs pages via generate-docs.py (andreessen skill/agent/commands; also reconciles claude-coach pages + dev-drift on code-reviewer/role-skill pages). - mkdocs build verified clean (only pre-existing unrelated relative-link warnings). https://claude.ai/code/session_01SF6MzfjHurZMt5JUFET9h3
…uctivity-plugin-PlOzn # Conflicts: # .claude-plugin/marketplace.json # .gemini/skills-index.json # docs/getting-started.md # docs/index.md # mkdocs.yml
Code Review — PR #737:
|
| Symlink | Before | After |
|---|---|---|
.codex/skills/review |
engineering-team/self-improving-agent/skills/review |
engineering-team/playwright-pro/skills/review |
.codex/skills/run |
engineering/autoresearch-agent/skills/run |
engineering/agenthub/skills/run |
.codex/skills/status |
engineering/autoresearch-agent/skills/status |
engineering/agenthub/skills/status |
This is a generator artifact (the sync script reordered duplicates), not a manual change. The swapped targets are real skills and the symlinks resolve correctly. However, users who relied on run/status pointing to autoresearch-agent will silently get agenthub variants. Worth a one-line note in the PR description for future auditability. No action needed now since it's already merged.
Observation — inner README version 1.0.0 vs plugin 2.8.4
The README footer says **Version:** 1.0.0 while plugin.json is 2.8.4. This is intentional (skill-level v1.0 vs repo-release v2.8.4 are different namespaces), but the handoff sibling inner README skips the version footer entirely to avoid this confusion. Consider aligning: either add (plugin v2.8.4) next to the skill version, or drop the footer like handoff does.
What's correct
- Gemini sync is complete and correct.
SKILL.mdsymlink in.gemini/skills/andreessen/matches the pattern of all sibling productivity skills (capture, handoff — each has only a SKILL.md symlink). Entry added to.gemini/skills-index.jsonwith correct categoryproductivity. - claude-coach gemini reconciliation is correct.
.gemini/skills/claude-coach/SKILL.mdnow symlinks toengineering/claude-coach/skills/claude-coach/SKILL.md, which resolves. Index entry added. - Unicode cleanup in marketplace.json. Converting
—/→/×to literal—/→/×is valid JSON and improves readability in editors without hurting any tooling. - New assets are high quality.
example_market_verdict.mdcorrectly models the anti-sycophancy voice (counterargument first, no hedging).forcing_question_worksheet.mdhas concrete recommended answers and attributes Sean Ellis's 40% gate to Ellis (not Andreessen) — this precision is correct and important.blank_3x5_card.mdis clean and matches the tool's output format. - plugin.json
skillspath follows canonical form ("./skills/andreessen"with./prefix).check_plugin_json.pywill pass. - vibe sync:
.vibe/skills/claude-skills/productivity/andreessensymlinks to the full skill directory (consistent with other vibe productivity entries).
Testing
| Check | Result |
|---|---|
| Symlinks resolve (gemini, vibe, codex) | PASS |
| Index counts consistent | PASS |
plugin.json schema (canonical ./ prefix) |
PASS |
| New asset files present | PASS (5 assets in directory) |
| README layout table covers all assets | FAIL — missing 4 of 5 |
| Python scripts present (3 scripts) | PASS |
Recommended follow-up: Update README.md layout table to include the 4 new assets. One 4-line addition.
Summary
Post-merge follow-ups from running the 8-phase plugin audit on
productivity/andreessen(shipped in #736). Two clean fixes:skills/andreessen/README.md. This was the one genuine documentation gap versus sibling skills; it lifts structure to 91.3/EXCELLENT and quality to 56.2.claude-coachentry the generator surfaced (claude-coach is a real engineering skill already in the codex index, just missing from gemini). Hermes needs no index change (dynamic discovery).Audit result: PASS WITH WARNINGS
The single warning (quality < 60) is a scorer-schema artifact:
skill_md_frontmatterandskill_md_sectionsexpect a title-caseName/Description/Features/Usage/Examplesschema that no Path-B productivity skill uses. On the same scorer, the already-merged siblings score lower (reflect 44.6, capture 46.4); handoff clears 60 only via a longer SKILL.md. I deliberately did not pad the doc with foreign sections to game the scorer.Note: the audit script's embedded "version must be 2.1.2 / strip extra fields" text is stale. Per
CLAUDE.md, version follows repo versioning (2.8.3) andattributionis an approved extension field. FollowedCLAUDE.md+ the livecheck_plugin_json.py(passes).Test plan
python3 engineering/skills/skill-tester/scripts/skill_validator.py productivity/andreessen/skills/andreessen→ EXCELLENT.codex/skills-index.jsonand.gemini/skills-index.jsonscripts/check_plugin_json.py --all→ andreessen OKhttps://claude.ai/code/session_01SF6MzfjHurZMt5JUFET9h3
Generated by Claude Code