The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local yadm and python commands to gather repository context. The implementation in scripts/gather_context.py uses structured argument lists with subprocess.run, which is a secure practice that correctly mitigates shell injection risks.
[SAFE]: The skill's operations are strictly confined to local dotfile management and PR generation. No evidence of data exfiltration, credential harvesting, or unauthorized file access was found.
[SAFE]: Regarding indirect prompt injection (Category 8), the skill ingests potentially untrusted data from git commit messages and code diffs. Although there are no explicit boundary markers or sanitization steps in scripts/gather_context.py, the risk is assessed as safe given the agent's constrained task of producing a specifically formatted PR summary and the lack of exploitable network or high-privilege tool access.

y-pr