The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash and Write tools to perform filesystem operations and archiving. It mitigates risk by requiring a dry-run summary and explicit user confirmation via the AskUserQuestion tool before taking action.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the local filesystem (directory names in .planning/phases/) which can influence agent behavior during archiving.
Ingestion points: Filesystem directory structure in .planning/phases/ (referenced in SKILL.md).
Boundary markers: None defined in the instructions to delimit directory names from commands.
Capability inventory: Bash, Write, Grep, Read (SKILL.md).
Sanitization: None specified for directory names before they are processed by shell tools.

gsd-cleanup