Agent Skills: Codebase Auditor

Paths: File paths (references/, ../ln-*) are relative to this skill directory.

Type: L2 Coordinator Category: 6XX Audit

Codebase Auditor

Mandatory Read

MANDATORY READ: Load references/evaluation_coordinator_runtime_contract.md, references/evaluation_summary_contract.md, references/evaluation_research_contract.md MANDATORY READ: Load references/audit_final_report_contract.md MANDATORY READ: Load references/codebase_audit_worker_boundaries.md MANDATORY READ: Load references/research_tool_fallback.md Conditional read: load references/epistemic_protocol.md only when source confidence or claim uncertainty affects issue validation.

Purpose

• audit runtime/codebase risks: security, delivery gates, dependency/reuse health, maintainability, dead code, diagnosability, concurrency, lifecycle, and config validation
• coordinate exactly ln-621 through ln-629
• require stack-aware research before scoring

Runtime Contract

Runtime family:

• evaluation-runtime

Identifier:

• codebase-audit

Phase order:

1. PHASE_0_CONFIG
2. PHASE_1_DISCOVERY
3. PHASE_2_RESEARCH
4. PHASE_3_DELEGATE
5. PHASE_4_AGGREGATE
6. PHASE_5_REPORT
7. PHASE_6_SELF_CHECK

Worker Set

• ln-621-security-boundary-auditor
• ln-622-build-delivery-gate-auditor
• ln-623-duplication-overabstraction-auditor
• ln-624-code-maintainability-hotspot-auditor
• ln-625-dependency-reuse-auditor
• ln-626-dead-code-pruning-auditor
• ln-627-diagnosability-auditor
• ln-628-concurrency-correctness-auditor
• ln-629-runtime-lifecycle-config-auditor

Worker Invocation (MANDATORY)

Host Skill Invocation: Skill(skill: "...", args: "...") is mandatory delegation.

• Claude: call the Skill tool exactly as shown.
• Codex: if no Skill tool exists, locate the named skill in available skills, read its SKILL.md, treat args as $ARGUMENTS, execute that skill workflow, then return here with its result/artifact.
• Do not inline worker logic or mark the worker complete without executing the target skill.

Use the Skill tool for delegated workers. Do not inline worker logic inside the coordinator.

TodoWrite format (mandatory):

• Resolve audit scope and build manifest
• Load codebase structure and stack
• Run best-practice research
• Delegate to domain audit workers
• Aggregate worker findings
• Generate final audit report and remediation plan
• Verify cleanup and self-check

Representative invocations:

Skill(skill: "ln-621-security-boundary-auditor", args: "{scope}")
Skill(skill: "ln-622-build-delivery-gate-auditor", args: "{scope}")
Skill(skill: "ln-623-duplication-overabstraction-auditor", args: "{scope}")
Skill(skill: "ln-624-code-maintainability-hotspot-auditor", args: "{scope}")
Skill(skill: "ln-625-dependency-reuse-auditor", args: "{scope}")
Skill(skill: "ln-626-dead-code-pruning-auditor", args: "{scope}")
Skill(skill: "ln-627-diagnosability-auditor", args: "{scope}")
Skill(skill: "ln-628-concurrency-correctness-auditor", args: "{scope}")
Skill(skill: "ln-629-runtime-lifecycle-config-auditor", args: "{scope}")

Workflow

Phase 0: Config

Start evaluation-runtime with required_research=true.

Phase 1: Discovery

Detect project type, stack, and applicability of audit workers.

Phase 2: Research

Mandatory research sources:

1. official docs or standards
2. MCP Ref
3. Context7 when framework docs matter
4. current web best-practice research

Phase 3: Delegate

Delegate applicable audit workers. Child workers must use evaluation-worker-runtime and emit evaluation-compatible summaries.

Phase 4: Aggregate

Merge runtime/codebase risk findings using references/codebase_audit_worker_boundaries.md. Read every worker report_path, normalize actions, deduplicate repeated issues, resolve worker conflicts, and validate each actionable problem against the research source order in references/evaluation_research_contract.md.

Phase 5: Report

Write .hex-skills/runtime-artifacts/runs/{run_id}/audit-report/ln-620--final-report.md per references/audit_final_report_contract.md. Include the remediation plan, source-backed validation for each confirmed issue, and cleanup note. Remove temporary worker markdown reports after consolidation. The evaluation-coordinator summary report_path must point to the final report only.

Phase 6: Self-Check

Required checks:

• [ ] research completed
• [ ] all applicable worker summaries recorded
• [ ] worker conflicts resolved with codebase_audit_worker_boundaries.md
• [ ] aggregation completed
• [ ] final remediation report written
• [ ] cleanup verified
• [ ] temporary worker markdown reports removed
• [ ] coordinator summary recorded

Summary Contract

Write summary_kind=evaluation-coordinator.

Definition of Done

• [ ] Evaluation runtime started
• [ ] Applicable workers selected
• [ ] Research completed
• [ ] All applicable worker summaries recorded
• [ ] Worker set remained exactly ln-621 through ln-629
• [ ] Worker conflicts resolved with codebase_audit_worker_boundaries.md
• [ ] Final report and remediation plan written
• [ ] Temporary worker markdown reports removed
• [ ] evaluation-coordinator summary written
• [ ] Runtime completed

Meta-Analysis

Optional reference: load references/meta_analysis_protocol.md only when the user asks for post-run meta-analysis or protocol-formatted run reflection.

When requested after the coordinator run, analyze the session per protocol section 7 and include the protocol-formatted output with the final codebase audit result.

References

• Workers: ../ln-621-security-boundary-auditor/SKILL.md, ../ln-622-build-delivery-gate-auditor/SKILL.md, ../ln-623-duplication-overabstraction-auditor/SKILL.md, ../ln-624-code-maintainability-hotspot-auditor/SKILL.md, ../ln-625-dependency-reuse-auditor/SKILL.md, ../ln-626-dead-code-pruning-auditor/SKILL.md, ../ln-627-diagnosability-auditor/SKILL.md, ../ln-628-concurrency-correctness-auditor/SKILL.md, ../ln-629-runtime-lifecycle-config-auditor/SKILL.md

Version: 5.0.0 Last Updated: 2025-12-23