The Agent Skills Directory

[SAFE]: The skill's primary function is to generate markdown documentation within the local docs/handoff/ directory. It does not perform network operations or exfiltrate data to external servers.
[COMMAND_EXECUTION]: Uses benign, read-only shell commands (date, git rev-parse, git log, git status) to collect context for the handoff document. These commands are restricted to gathering local environment metadata (timestamp, branch name, and commit history).
[DATA_EXFILTRATION]: Includes explicit safety instructions for the AI agent to avoid including sensitive data: "Keep secrets out. Test-mode IDs and public config are fine; real API keys, tokens, and passwords never go in a committed doc."
[PROMPT_INJECTION]: Instructions focus on task summarization and context preservation. No attempts to override safety filters or hijack agent behavior were found.

handoff-torch