The Agent Skills Directory

[SAFE]: The skill operates primarily as a read-only documentation assistant. Its instructions explicitly forbid modifications to shared knowledge, Kanban boards, or project manifests, ensuring that it cannot be used to perform unauthorized state changes.
[SAFE]: Write capabilities are strictly limited to a local '.feedback' directory. This process is protected by multiple checks: it requires an explicit user request, a manifest setting ('feedback.enabled'), and verification that the directory is excluded from version control (via .gitignore). This prevents accidental exposure or exfiltration of the recorded notes.
[SAFE]: The skill incorporates clear boundary management, instructing the agent to stop after providing prompt suggestions for other skills. It refuses to continue execution into write-capable skills within the same invocation, maintaining a human-in-the-loop requirement for any state-changing actions.
[SAFE]: No evidence of network communication, remote script execution, obfuscated content, or unsafe dependency usage was found across the skill's markdown files and configuration.

knowledge-assistant