The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Python scripts to interact with the local environment for gathering project metadata.
Evidence: scripts/create_handoff.py and scripts/check_staleness.py execute Git commands using the subprocess module. These calls are implemented using argument lists rather than shell strings, which is the recommended secure pattern to prevent command injection.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and processing generated markdown documents from the local filesystem, which constitutes a potential surface for indirect prompt injection.
Evidence: The skill reads files from the .claude/handoffs/ directory during the RESUME Workflow. While these documents are used to transfer context between agent sessions, the skill includes a validate_handoff.py script and a resume-checklist.md to help the agent and user verify the integrity and safety of the transition.
[SAFE]: The skill includes proactive security measures to prevent data leakage.
Evidence: The scripts/validate_handoff.py script performs regular expression scanning for over a dozen sensitive patterns, including OpenAI API keys, GitHub tokens, and private keys. It is designed to block the finalization of a handoff if secrets are detected, ensuring that sensitive credentials are not stored in plain-text markdown files.

session-handoff