repo-wide-dead-code-sweep
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from the repository, including source code, database catalogs, and external analyzer reports, which could potentially contain malicious instructions aimed at influencing the agent's deletion decisions.
- Ingestion points: The skill reads repository source code,
knip.jsonanalyzer reports, translation files (e.g.,locales/*.json), and database metadata (SKILL.md Steps 1-3). - Boundary markers: The instructions mandate the creation of a 'four-part dossier' for every candidate and require explicit owner approval and 'positive controls' to validate search results before any action is taken.
- Capability inventory: The skill utilizes shell execution (git, grep, jq), SQL catalog queries, and file system modifications (micro-commits for code deletion).
- Sanitization: The provided recipes use
jqandsedto extract specific keys and patterns, providing basic structured parsing for data ingestion.
Audit Metadata