The Agent Skills Directory

[SAFE]: The skill implements legitimate session management logic intended to preserve work and transition to a handoff state.
[COMMAND_EXECUTION]: Uses standard development tools including git status, git diff, and git commit to inventory and secure the working tree. These operations are limited to the local project environment and follow existing session authorization policies.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and synthesize status reports from subagents (workers). While this represents a data ingestion surface, the source is internal to the agent's own orchestrated session.
Ingestion points: Subagent reports in SKILL.md (Worker Wind-Down section).
Boundary markers: Absent.
Capability inventory: Local file operations and Git commands in SKILL.md.
Sanitization: Absent.

wrap-up