prune-codebase
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted codebase data, including source code, documentation, and agent directives, which could contain adversarial instructions intended to influence the agent's pruning or modification behavior.
- Ingestion points: The skill ingests data from subsystem surveys,
AGENTS.md,CLAUDE.md, Architecture Decision Records (ADRs), and various documentation files across the repository. - Boundary markers: The instructions do not define explicit delimiters or boundary markers to differentiate untrusted codebase content from the agent's own operational instructions.
- Capability inventory: The skill possesses significant capabilities, including file system access for reading and writing, shell command execution for repository surveying and test runs, and the ability to exercise application entry points.
- Sanitization: The instructions do not specify technical sanitization or escaping for external content, though they include strong process-level mitigations, such as mandates to preserve security properties and public contracts, and the requirement to handle security remediation as separate, user-approved tasks.
Audit Metadata