The Agent Skills Directory

[COMMAND_EXECUTION]: The skill runs a local Python script (.sentinel/compaction/compact.py) to classify session candidates. This execution is limited to project-specific automation and uses static paths within the repository's infrastructure.
[PROMPT_INJECTION]: The skill processes untrusted content from modified repository files and session context to generate new rules. This constitutes an indirect prompt injection surface. However, the risk is effectively mitigated by the skill's design, which explicitly requires user review and approval via AskUserQuestion before any files are written or promotion logic is executed.
[SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration were detected. The skill follows best practices for secret management by not requesting sensitive environment variables and focuses strictly on project metadata management.

progress