The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes system commands such as grep and find to scan the codebase for deprecation markers and legacy file naming conventions. It also executes language-specific build and test runners including bun, tsc, eslint, pytest, go test, and cargo test to verify code integrity after automated deletions.
[DATA_EXFILTRATION]: The skill reads repository configuration and environment files, specifically checking .env, .env.staging, and .env.production. This is used to determine if feature flag branches are static and can be safely removed, representing local data access for code analysis purposes.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks as it ingests and processes untrusted data from the repository being analyzed.
Ingestion points: Data enters the agent context via grep and find operations on the repository files.
Boundary markers: None are present to distinguish between legitimate code and maliciously injected markers.
Capability inventory: The skill has the capability to delete files (rm), modify source code blocks, and commit changes to the git repository.
Sanitization: The skill relies on caller-verification logic (ensuring zero references) and successful test suite execution to prevent erroneous or malicious deletions.

cleanup-legacy