skills

Skill Quick Reference

Explicit invocation only — host-neutral. disable-model-invocation: true enforces this in Claude Code and in VS Code Copilot, which reads .claude/skills/ and honours that key (verified against VS Code Agent Skills docs 2026-09-02). Hosts beyond those two are unverified, and a key cannot carry the reasoning anyway, so the rule is stated here as well.

Nothing here is destructive — this is a roster, not a gate — but a full skill listing is a poor answer to a question about the work, so render it when the user asked for it.

A discoverable, always-current reference for every skill in this bundle. Reads directly from skills/*​/SKILL.md frontmatter so the listing can never drift from reality — the same file Claude Code parses to register each skill.

This skill itself does NOT modify code. Pure read-and-render.

Worktree preflight — in a linked git worktree the synced tooling tree scripts/.claude-skills/ is absent — it is gitignored, so git worktree add does not populate it, and every command below that uses it dies on a bare MODULE_NOT_FOUND. Run npm run skills:hydrate first. If this repo defines no such script, references/worktree-preflight.md holds the entry to add — ask the user first, since it edits their package.json.

---

Step 0 — Pick the mode

Input	Mode
/skills (no argument)	List all — compact table of every skill
/skills <skill-name>	Detail — full description, triggers, usage for one skill
/skills --search "<term>"	Filter — show only skills whose name/description/triggers/usage contain the term (case-insensitive)

If <skill-name> doesn't match an existing skill, the helper suggests near-matches. Skill names are the directory names under skills/ (e.g. audit-code, explain, persona-test).

---

Step 1 — Invoke the helper

# List all skills
node scripts/skills-help.mjs

# Detail for one skill (SKILL_NAME = a directory name under skills/, e.g. audit-code)
node scripts/skills-help.mjs SKILL_NAME

# Filter by search term
node scripts/skills-help.mjs --search "<term>"

The helper auto-detects whether to render a compact table, a per-skill detail block, or a filtered list, and emits Markdown by default. Output is read-only — no files written unless --out <path> is passed.

---

Step 2 — Render the result

The helper output is already Markdown. Pass it through verbatim — do NOT re-summarise (the user asked for the reference; don't paraphrase).

If the helper exits non-zero with "skill not found", surface the suggestion line as-is and stop.

---

Output format reference

Compact list

# Available skills (N total)

| Skill | One-liner |
|---|---|
| `/audit-code` | Iteratively audit code against a plan with GPT + Gemini final gate. |
| `/brainstorm` ? | Multi-LLM concept-level brainstorming. |
| ... |

? = `disable-model-invocation: true` — Claude Code and VS Code Copilot will not
auto-trigger it; invoke explicitly via `/<name>`. Each locked skill also states
the rule in its body, for hosts that ignore the key.

Detail (e.g. /skills explain)

# /explain

Explain WHY a piece of code is structured the way it is.

**Triggers on:**
- why is this
- explain this code
- ...

**Usage:**
\`\`\`
/explain <file>                Explain the file's purpose + history
/explain --history "<topic>"   "Did we already solve this?" — cross-source search
\`\`\`

**Full SKILL.md:** `skills/explain/SKILL.md`

---

When to use this vs. typing /

• Type / in Claude — for autocomplete on a skill you already know the name of (fastest single-skill lookup)
• Run /skills — when you want to discover what's available, refresh your memory on a skill's options, or share a quick reference with a teammate (the output is copy-pasteable)
• Open the SKILL.md directly — when you need the full "how it works" body (this skill only renders the frontmatter contract)

---

Sources

• skills/*​/SKILL.md frontmatter — the canonical truth read by both Claude Code AND this helper. If /skills shows wrong info, the SKILL.md itself is wrong (not the helper).
• .claude/skills/ mirror is intentionally ignored — that's the generated copy, not authoritative.

Cost

Zero external calls. Pure local file scan. Sub-100ms even with 50+ skills.

---

Reference files

File	Summary	Read when
references/worktree-preflight.md	Worktree preflight remedy — the skills:hydrate npm script to add when a linked worktree lacks the tooling tree.	A command dies with MODULE_NOT_FOUND in a linked worktree AND this repo has no skills:hydrate script.