The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the subprocess module to execute informational git commands (such as git status, git log, and git branch) to track work progress and reconcile repository state between sessions. These commands are static and do not involve untrusted input execution.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and follow 'Next Recommended Action' and 'Resume Instructions' from handoff files stored in local directories. This creates an inherent surface for indirect instruction processing, which the skill manages through clear data structure boundaries and a mandatory validation process.\n
Ingestion points: scripts/session_transfer_lib.py and scripts/resume_handoff.py read markdown handoff files from agent-specific directories like .codex/handoffs/.\n
Boundary markers: Metadata is segregated into YAML code blocks, and the content is organized using standard Markdown headers to delimit instructions from data.\n
Capability inventory: The skill can write files to project-local agent directories and run informational git commands.\n
Sanitization: The scripts/validate_handoff.py utility provides a security gate by scanning handoff text for secrets (API keys, tokens, passwords) using regular expressions and blocking the finalization of handoffs that fail these safety checks.\n- [DYNAMIC_EXECUTION]: The scripts scripts/check_staleness.py and scripts/create_handoff.py use the __import__ function to dynamically load standard library modules os and re. This is used for portable module loading and does not indicate malicious runtime code generation or obfuscation.

session-transfer