ask-parallel-auditor

An orchestrator skill designed to bypass single-agent context limits by chunking repositories and spawning parallel subagents to conduct massive audits.

Purpose

When executing a repository-wide security scan, architectural audit, or massive refactor, a single AI agent will quickly blow out its context window if it tries to read hundreds of files sequentially.

ask-parallel-auditor implements the Hierarchical Multi-Agent System (HMAS) pattern. It acts as the Strategy Layer Orchestrator. It does not read the files itself. Instead, it chunks the repository and spawns multiple, isolated subagents (like ask-owasp-security-review).

<critical_constraints> ❌ STRICTLY NO attempting to audit the files yourself sequentially. Your context window will fail. ❌ STRICTLY NO modifying source code or logic. ✅ MUST explicitly chunk the target files into logical groupings. ✅ MUST delegate the auditing task to parallel background workers/subagents. ✅ MUST aggregate the structural output from the subagents into a unified PARALLEL_AUDIT_REPORT.md. </critical_constraints>