full-audit
Automated, multi-dimensional audit suite that orchestrates agents to produce a consolidated quality report.
Install
mkdir -p .claude/skills/full-audit && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/11342" && unzip -o skill.zip -d .claude/skills/full-audit && rm skill.zipInstalls to .claude/skills/full-audit
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Run a unified audit across all quality dimensions — architecture, code quality, testing, security, docs, and release readiness. Consolidates findings from multiple agents and scripts into a single deduplicated report.Key capabilities
- →Run multi-wave audits
- →Deduplicate findings
- →Generate consolidated reports
- →Track resolution status
How it works
It orchestrates parallel audit waves and uses a 3-pass deduplication engine to merge findings into a single report.
Inputs & outputs
When to use full-audit
- →Pre-release health check
- →Technical debt assessment
- →Security and documentation audit
- →Architectural review
About this skill
/full-audit
Unified project audit that orchestrates scripts and agents in waves, deduplicates findings, and produces a consolidated report.
Usage
/full-audit # standard profile (default)
/full-audit --profile quick # fast, no Gradle
/full-audit --profile deep # everything including beta-readiness
/full-audit --verbose # detailed inline findings
/full-audit --output report.md # save report to file
/full-audit --project-root PATH # target project (default: cwd)
/full-audit --skip coverage,sbom # skip specific checks
Profiles
| Profile | Waves | Gradle | Agents | Estimated Time |
|---|---|---|---|---|
quick | 1-2 | No | No beta-readiness | ~5 min |
standard | 1-3 | Yes (coverage) | No beta-readiness | ~15 min |
deep | 1-4 | Yes (full) | All | ~45 min |
Procedure
Step 1: Resolve Configuration
- Parse arguments:
--profile(default: standard),--verbose,--output,--project-root,--skip - Read
skills/full-audit/profiles.jsonfor wave definitions - If
$PROJECT_ROOT/l0-manifest.jsonexists, check forfull_auditoverrides:extra_agents: additional agents to includecustom_profiles: profile overridesskip_checks: checks to always skip
Step 2: Execute Waves
Delegate to the full-audit-orchestrator agent, passing:
- Profile name and wave configuration
- Project root path
- Skip list
- Verbose flag
The orchestrator will:
- Execute each wave's checks in parallel where possible
- Collect findings using the
<!-- FINDINGS_START -->/<!-- FINDINGS_END -->protocol from agents - Run the 3-pass deduplication engine on all collected findings
- Track resolution against previous
findings-log.jsonlentries
Step 3: Produce Report
The orchestrator produces the consolidated report in this format:
================================================================
FULL AUDIT REPORT
{Project} -- {Date} -- Profile: {profile}
================================================================
HEALTH: {CRITICAL|WARNING|HEALTHY} ({N} findings: X CRIT, Y HIGH...)
CRITICAL [security] OAuth tokens in SQLite plaintext
CRITICAL [architecture] java.time.* in commonMain (4 files)
HIGH [code-quality] CancellationException swallowed (3 repos)
MEDIUM [documentation] 3 pattern docs with stale versions
...
Checks: {N}/{M} | Duration: {T} | Deduped: {D} findings merged
ACTION ITEMS (priority order):
1. [CRITICAL] Encrypt token storage -- Art. 32 compliance
2. [CRITICAL] Replace java.time with kotlinx-datetime
3. [HIGH] Add rethrow in catch blocks -- /auto-cover can help
================================================================
Step 4: Persist & Output
- Append findings to
.androidcommondoc/findings-log.jsonlusing the findings-append helper - Append a
full_auditevent to.androidcommondoc/audit-log.jsonlusing audit-append helper - If
--outputspecified, write report to file - Display report to user
Constraints
- NEVER re-run Gradle if
--profile quickis selected - NEVER run beta-readiness-agent unless
--profile deep - Respect
--skiplist -- silently skip named checks - All agents must emit findings using the structured protocol
- The report must be <=200 lines (truncate detail section if needed, keep action items)
When not to use it
- →Single-file linting
Prerequisites
Limitations
- →Report limited to 200 lines
How it compares
It consolidates disparate quality checks into one unified report rather than running isolated tools.
Compared to similar skills
full-audit side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| full-audit (this skill) | 0 | 3mo | No flags | Intermediate |
| code-review-checklist | 3 | 27d | No flags | Beginner |
| pr-review | 0 | 1mo | No flags | Advanced |
| reviewing-code | 21 | 8mo | No flags | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
code-review-checklist
vudovn
Code review guidelines covering code quality, security, and best practices.
pr-review
microsoft
Multi-dimensional review of a PR or feature branch in the microsoft/winappcli repo. Activate when a contributor asks to "review my PR", "review my changes", "vet my branch before pushing", "do a full review", "PR review", "review this feature", or similar. Fans out parallel sub-agents covering secur
reviewing-code
CaptainCrouton89
Systematically evaluate code changes for security, correctness, performance, and spec alignment. Use when reviewing PRs, assessing code quality, or verifying implementation against requirements.
cookbook-audit
anthropics
Audit an Anthropic Cookbook notebook based on a rubric. Use whenever a notebook review or audit is requested.
pr-review
pytorch
Review PyTorch pull requests for code quality, test coverage, security, and backward compatibility. Use when reviewing PRs, when asked to review code changes, or when the user mentions "review PR", "code review", or "check this PR".
fix-review
sickn33
Verify fix commits address audit findings without new bugs