The state 88 of 110 skills actively used, wired, or project-scoped; 22 idle global skills lack wiring (correctly unwired for manual invocation when needed).

Fix wiring No fixes required — all active and wired skills fire correctly; no duplicates or name collisions detected.

Combine Merge baoyu-compress-image, baoyu-format-markdown, baoyu-url-to-markdown into baoyu-utils — all three are personal content-utility scripts with identical scope. Verify whether business-intelligence and dataviz serve a shared "data-insights" job (distinct from plugin data:* skills) or remain separate; if shared, merge; if not, keep both. Do NOT merge strategy skills: blue-ocean-strategy + good-strategy-bad-strategy solve different problems (market creation vs. strategy evaluation), as do jobs-to-be-done + mom-test (motivation discovery vs. validation testing).

Keep All active (immigration-pipeline, brainstorming, artifact-design, subagent-driven-development, finishing-a-development-branch, run, claude-api, writing-plans), wired (caveman, grill-me, to-issues, git-guardrails-claude-code, shadcn), on-hold (ceo-sim-ui-design-system, pixel-game-uiux, ui-ux-pro-max), and 71 project-local skills untouched.

The state

106 skills: 9 active, 5 wired (auto-fired per session invisibly), 3 on-hold, 71 project-local; 18 idle & unwired. Setup is sound—no collisions detected—but 5 strategy frameworks installed 2026-07-06 (good-strategy-bad-strategy, jobs-to-be-done, mom-test, pricing-strategy, market-sizing) sit unused and share an underlying job.

Fix wiring

No duplicates or name collisions. grill-with-docs: Likely overlaps with grill-me (wired globally for complex-task clarification). If distinct, add a trigger phrase to a project CLAUDE.md; otherwise remove it.

Combine

• strategy-frameworks (merge good-strategy-bad-strategy + jobs-to-be-done + mom-test + pricing-strategy + market-sizing): All five apply frameworks to strategic decisions—same meta-job, different tools. Unify into one skill with context-aware suggestions (pricing context → pricing-strategy; market context → market-sizing; strategy evaluation → good-strategy-bad-strategy). Reduces friction and wiring overhead.

Keep

ACTIVE: immigration-pipeline, brainstorming, artifact-design, finishing-a-development-branch, research-batch, writing-plans, subagent-driven-development, diagnose. WIRED: caveman, git-guardrails-claude-code, grill-me, shadcn, to-issues. ON-HOLD: ceo-sim-ui-design-system, pixel-game-uiux, ui-ux-pro-max. PROJECT-LOCAL: all 71 skills. Untouched.

The state

Setup is intentional: 88 skills in active use (9 invoked, 5 wired-but-uncounted, 3 on-hold, 71 project-local). 16 idle-global are recent strategic tools (wondelai, 2026-07-06) or personal utilities—unwired or not yet triggered, not dead.

Fix wiring

• grill-with-docs (idle) vs grill-me (wired via CLAUDE.md for complexity). Verify they're distinct—if redundant, keep grill-me; if complementary, add grill-with-docs trigger to CLAUDE.md or remove.
• Strategy trio (good-strategy-bad-strategy, jobs-to-be-done, mom-test): Installed 3 weeks ago, zero invocations. Wire into CLAUDE.md to auto-recommend on business/proposal work, or explicitly mark "manual-invoke-only."

Combine

None—strategy frameworks are complementary methodologies, not the same job. Personal utilities (baoyu-*) intentionally separated by function.

Keep

Active 9 (immigration-pipeline 166 uses). Wired 5 (caveman, git-guardrails-claude-code, grill-me, shadcn, to-issues). On-hold 3. Project-local 71 all untouched.

The state: 85 of 104 skills are properly placed — 5 active (immigration-pipeline dominates at 154 uses), 6 auto-wired every session (caveman, system-design, to-issues, etc.), 3 paused-project-expected, 71 project-local. The 19 idle global skills are deliberate installs, not dead weight; de-slop is mis-wired; the rest are conditional-trigger utilities the user hasn't invoked yet.

Fix wiring: - de-slop: installed 2026-07-06 but never wired despite being a behavioral modifier (like caveman mode). If meant to auto-humanize responses, add to global CLAUDE.md as: "Response humanization: caveman mode for tight prose; de-slop call on review requests." Otherwise, clarify in the skill description when to invoke it manually.

Combine: - None. The 19 idle skills address distinct use cases: Baoyu utilities (compression, transcription, translation, URL→MD), strategy frameworks (good-strategy, JTBD, mom-test), BI/reporting (business-intelligence, report-generator), research (tavily, zoom-out, grill-with-docs), design (html-diagram, tailwind-design-system, improve-codebase-architecture). Merging by surface topic (e.g., "research" buckets) would blur triggers and weaken signal. Keep them separate for clear invocation.

Keep: All 5 active, all 6 wired (caveman, git-guardrails-claude-code, grill-me, shadcn, system-design, to-issues), all 3 on-hold CEO Sim design skills, and all 71 project-local skills across auto-social-posting, ClientCo, ai-learning-path, etc. — untouched.

The state: 43 skills across projects; 8 actively invoked in 30 days (immigration-pipeline 117×, research-batch 4×, impeccable 3×, and 5 others 1× each), 6 wired into rules (applied every session), 3 on-hold, 5 project-local—24 idle are mostly deliberate utility/framework tools, not dead.

Fix wiring: Remove taste-skill (shadows design-taste-frontend on invocation name). Keep the fuller skill; delete the shadow file to stop the collision. No hooks need changes.

Combine: Merge business-analytics-reporter + report-generator into single "business-reporting" skill (same job, different domains—keep the broader name). Keep good-strategy-bad-strategy, jobs-to-be-done, mom-test separate for now (distinct methodologies, distinct triggers).

Keep: Active—immigration-pipeline, research-batch, impeccable, writing-plans, dataviz, artifact-design, brainstorming, claude-api. Wired—caveman, diagnose, git-guardrails-claude-code, grill-me, shadcn, to-issues. All on-hold & project-local skills. Do not touch.