The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands including git mv, mv, and rm to archive or delete files within the project repository as described in the behavior section of SKILL.md.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and acts upon repository context that may be controlled by an attacker.
Ingestion points: The skill accepts user_instruction, file_list, and context as inputs, specifically utilizing findings from other tools like onboard-repo to identify targets (SKILL.md, agent.yaml).
Boundary markers: The skill explicitly requires user confirmation of the proposed plan before execution, acting as a manual checkpoint (SKILL.md).
Capability inventory: The skill can move and remove files using git mv, mv, and rm across the repository (SKILL.md).
Sanitization: There is no automated sanitization of the input context; the skill relies entirely on the user to review and approve the suggested operations.

prune-content