The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to execute Git commands for repository state verification (e.g., git rev-parse HEAD, git status --porcelain). These are standard operations for development tools.
[COMMAND_EXECUTION]: In validate_packet.py, the skill can execute 'command probes' and 'receiver checks'. These executions are strictly gated by a user-controlled allowlist in .claude/session-boundary.json. The validator specifically rejects any command that is not explicitly authorized in the local configuration file, preventing arbitrary command injection from the packet content itself.
[DATA_EXPOSURE]: The skill scans for secrets (e.g., API keys, private keys) within the handoff packet using regex patterns. This is a safety feature intended to prevent the accidental inclusion of credentials in session transcripts or handoff files.
[REMOTE_CODE_EXECUTION]: No remote code execution or external downloads were detected. All scripts (snapshot_state.py, validate_packet.py) use Python standard library components and operate only on local repository files and Git state.

im-down