The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill suggests installing companion utilities using npx skills add crissmoldovan/agent-skills. These resources are managed by the skill's author and represent official extensions.
[COMMAND_EXECUTION]: Analysis is performed using standard shell utilities, including git grep, git ls-files, and wc, which are used to search and index files within the target repository.
[INDIRECT_PROMPT_INJECTION]: The skill inherently processes untrusted data from the codebase being investigated. It includes architectural safeguards such as agent isolation, 'independence' rules for children, and a formal evidence-reconciliation process to verify claims and mitigate the risk of untrusted content influencing the final result.
Ingestion points: Reads arbitrary files from the repository checkout during search and walk phases.
Boundary markers: Explicitly mandates 'withholding parent context from children' and establishing 'closure over boundaries' for reachability claims.
Capability inventory: Uses Read, Write, Grep, Glob, and Bash tools to analyze the repository.
Sanitization: Relies on a multi-stage reconciliation table and adversarial review (in deep mode) to validate findings against source artifacts.

investigate-codebase