The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands to verify the status of work units and to finalize archival, specifically calling node .agents/scripts/cooking.mjs status and the git-commit utility.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect injection surface as it makes operational decisions (specifically the deletion of directories) based on data read from external files and script outputs.
Ingestion points: The skill ingests status data from the output of the cooking.mjs script and from markdown files located in the reviews/ and .agents/ directories.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present for the data being processed.
Capability inventory: The skill has the capability to delete entire local directories (.agents/cooking/) and initiate git commits.
Sanitization: The workflow includes a series of logical requirements and status checks (e.g., verifying specific field values like '可归档：是') that act as a validation layer before any destructive action is taken.

archive