---

description: "Run a preset suite of audits in parallel" argument-hint: "--preset <name>"

Audit Suite Orchestrator

Runs a named preset — a curated list of other /audit-* skills — by fanning them out as background agents in parallel, then merges their reports into one summary. This skill owns no audit logic of its own; it only sequences and aggregates. Shared protocol (project layout, severity, dedup, report format) lives in .claude/commands/_audit-common.md and .claude/commands/_audit-severity.md — not repeated here.

Every audit referenced below is a live skill at .claude/commands/audit-<name>/SKILL.md, invoked as /audit-<name>. The full subsystem + meta set (12): pipeline, nes-hardware, patterns, exporters, dpcm, arranger, mappers, tempo, performance, safety, tech-debt, regression. (/audit-incremental is the delta meta-audit; /audit-publish is post-processing; /audit-sync retires stale prose in these very skill files after a fix lands — none of the three are ever part of a preset.)

Preset Index

| Preset | When | Audits | |--------|------|--------| | quick | after any change, < 10 min | incremental | | pre-release | before tagging a version | pipeline · nes-hardware · mappers · regression | | comprehensive | monthly / pre-milestone | all 12 | | tech-debt-deep | after a milestone closes | tech-debt · incremental | | audio-correctness | after APU/channel/pitch/envelope work | nes-hardware · tempo · exporters | | compression-deep | after pattern/compression work | patterns · exporters · regression | | rom-deep | after mapper / project-builder / compiler work | mappers · pipeline · nes-hardware | | arranger-deep | after arranger / voice-allocation work | arranger · nes-hardware · tempo | | dpcm-deep | after DPCM / drum work | dpcm · exporters | | pipeline-deep | after main.py / stage-contract work | pipeline · patterns · exporters · safety | | songbank-deep | after song build / jukebox work | pipeline · exporters · mappers · nes-hardware |

Presets

--preset quick

Fast sanity check after a change (< 10 min):

1. /audit-incremental --commits 5

--preset pre-release

Run before tagging a release — the stages that decide whether a ROM boots and plays:

1. /audit-pipeline
2. /audit-nes-hardware
3. /audit-mappers
4. /audit-regression

--preset comprehensive

Full coverage (longest — run monthly or before a major milestone). Every subsystem + both meta audits:

1. /audit-pipeline
2. /audit-nes-hardware
3. /audit-patterns
4. /audit-exporters
5. /audit-dpcm
6. /audit-arranger
7. /audit-mappers
8. /audit-tempo
9. /audit-performance
10. /audit-safety
11. /audit-tech-debt
12. /audit-regression

--preset tech-debt-deep

Surface accumulated debt (run after a milestone closes, before opening the next):

1. /audit-tech-debt
2. /audit-incremental --commits 30

--preset audio-correctness

After APU register / channel / pitch-table / envelope work — the path from a note to a register write:

1. /audit-nes-hardware
2. /audit-tempo
3. /audit-exporters

--preset compression-deep

After pattern-detection or compression changes — round-trip integrity is the headline risk:

1. /audit-patterns
2. /audit-exporters
3. /audit-regression

--preset rom-deep

After mapper, project-builder, or CC65-compiler changes:

1. /audit-mappers
2. /audit-pipeline
3. /audit-nes-hardware

--preset arranger-deep

After arranger / role-analysis / voice-allocation / arpeggiation changes:

1. /audit-arranger
2. /audit-nes-hardware
3. /audit-tempo

--preset dpcm-deep

After DPCM sample / drum-mapping changes:

1. /audit-dpcm
2. /audit-exporters

--preset pipeline-deep

After main.py dispatch or stage-contract changes:

1. /audit-pipeline
2. /audit-patterns
3. /audit-exporters
4. /audit-safety

--preset songbank-deep

After song build / multi-song jukebox changes (#30/F-13). This feature is one code path spread across four subsystems, and no single-song test or ROM exercises any of it — the engine additions are .ifdef JUKEBOX_BUILD-gated, so an ordinary regression run passes green while the jukebox path is broken. Its first audit pass found defects that produced zero working ROMs at any bank size. Run all four; findings in one almost always have a counterpart in another (the exporter emits what the engine reads):

1. /audit-pipeline — Dimension 8 (bank load → per-song frames → capacity pre-flight)
2. /audit-exporters — Dimension 9 (export_song_bank_bytecode, song_table, prefixes)
3. /audit-mappers — Dimension 7 (the song_count is not None → JUKEBOX_BUILD gate)
4. /audit-nes-hardware — Dimension 11 (.ifdef JUKEBOX_BUILD engine routines)

Execution

1. Parse the --preset argument from $ARGUMENTS. If unknown, list the preset index above and stop.
2. mkdir -p /tmp/audit.
3. Launch each audit in the preset as a background agent, max 3 concurrent. The audits are independent — they read the tree and write distinct reports — so they fan out in parallel with no ordering dependency.
4. Each audit writes its own report to docs/audits/AUDIT_<TYPE>_<TODAY>.md (per _audit-common finalization).
5. When all complete, produce a combined summary:

# Audit Suite Summary — <preset> — <date>

| Audit | Findings | CRITICAL | HIGH | MEDIUM | LOW | Report |
|-------|----------|----------|------|--------|-----|--------|
| pipeline | 3 | 0 | 1 | 2 | 0 | AUDIT_PIPELINE_... |
| ...   | ... | ... | ... | ... | ... | ... |

Total: X findings (C critical, H high, M medium, L low)

6. If any CRITICAL findings exist, warn prominently at the top of the summary.
7. For each report that has findings, suggest: /audit-publish docs/audits/AUDIT_<TYPE>_<TODAY>.md