The Agent Skills Directory

[DATA_EXPOSURE]: The skill scans local configuration files, project journals, and activity logs to perform its audit. This access is restricted to the local filesystem and is necessary for its primary administrative functions.
[PROMPT_INJECTION]: As the skill ingests metadata and documentation from other agents, it possesses an indirect prompt injection surface. However, its strictly proposal-based architecture and requirement for human approval for any archival actions mitigate potential exploitation risks.
[SAFE]: The skill documentation establishes clear boundaries that protect 'core' infrastructure agents from being proposed for removal, ensuring ecosystem stability.
[SAFE]: Destructive actions are not supported by the skill's code; it only generates reports and delegates execution to either architectural agents (merges) or manual user approval (archiving).

prune