The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local code analysis tools (such as knip, jscpd, ruff, or vulture) and git commands to perform and verify refactoring work within the project repository.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves scanning the project workspace to identify cleanup candidates, which establishes a potential surface for indirect injection if malicious instructions are embedded in source code comments or strings.
Ingestion points: The entire workspace codebase is inspected during the 'sweep' and 'rotation' phases described in preset.md.
Boundary markers: There are no explicit markers or instructions provided to the agent to distinguish between inert code and potential embedded instructions in the analyzed files.
Capability inventory: The skill has authority to perform local file edits, run verification tests, and execute git commands for local commits.
Sanitization: The skill does not define specific sanitization or validation logic for the data ingested from the repository files.
[DYNAMIC_EXECUTION]: The skill includes instructions to dynamically generate a temporary duplication detector script if a standard one is not present in the user's environment, which is then executed locally to provide a baseline for code cleanup (preset.md).

loop-converge