The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes common shell utilities including find, grep, jq, and wc to inspect local environment configuration and project transcripts.
[DYNAMIC_EXECUTION]: Step 6.5 contains an embedded Python script that performs directory analysis to identify orphaned memory files. It also references a local linting script at ~/.claude/scripts/harness-lint.py.
[INDIRECT_PROMPT_INJECTION]: The skill processes content from potentially untrusted files like CLAUDE.md and session transcripts. Ingestion points: Configuration files and project transcript logs. Boundary markers: None are applied during file analysis. Capability inventory: Local filesystem read access and shell command execution. Sanitization: None; the tool relies on raw text processing for auditing.

setup-audit