From orchestration-studio
Run a deep codebase audit by partitioning it into areas and delegating each to a parallel worker agent with a findings-only mandate, then consolidating the reports into one prioritized document. Use when the user asks for a thorough audit, wants to map every component for a class of bugs, or wants to verify code health across a large codebase.
How this skill is triggered — by the user, by Claude, or both
Slash command
/orchestration-studio:audit-swarmThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
1. **Findings-only mandates.** Auditors REPORT, they don't fix. Fixing during
AUDIT (findings only, NO fixes) — area: <LAYER>.
Context: <the confirmed bug family, 2-3 lines, with PR references>.
Work in <worktree> on a read-only branch from origin/main.
MANDATE:
(a) <enumerable inventory task — "classify EVERY X by Y">
(b) <second dimension>
(c) <standard checks for this layer: dead code / perf / validation consistency>
For each finding: file:line, class, severity (CRITICAL/HIGH/MEDIUM/LOW),
hypothetical repro, 1-line suggested fix.
Write the report to <path>/audit-<area>.md and reply with the severity count.
The enumerable inventory framing ("all 20 forward types", "every client.* call site", "all 63 handlers") is what forces completeness — auditors given open-ended scope sample; auditors given a checklist enumerate.
Require an explicit per-severity COUNT at the end of each report — it makes consolidation and progress tracking mechanical.
audit-<area>.md reports..dev-studio/audit-report.md (or equivalent): summary table,
findings by severity, then cross-cutting themes — the most valuable
output. Real example: 21 separate HIGH findings collapsed into ONE
architectural fix ("storage writes are fire-and-forget with no outbox");
another 10 collapsed into one decision ("entity X lives in per-node local
storage but is consumed as global — move to shared DB or require explicit
node targeting everywhere")."Open issue count" is a lie after a big merge wave. Before dispatching a fix wave over an old backlog, run this variant — the target is the issue tracker against current HEAD, not the code blind:
gh issue view, check the code at HEAD, emit
{issue, verdict: FIXED | STILL-VALID | PARTIAL, evidence, fixing_pr?}.Skipping this means re-implementing already-fixed issues (wasted waves) and missing the regressions; the verification doubles as regression detection.
npx claudepluginhub rp4ri/orchestration-studioProvides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.