The Agent Skills Directory

[SAFE]: The skill provides a structured workflow for session management and documentation without introducing malicious patterns or dangerous command execution.
[SAFE]: Explicit security guardrails are included in the instructions, specifically requiring the agent to replace secrets (keys, tokens, passwords) and personal data with [redacted] before writing them to disk.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface in that it ingests session history and project data to generate summaries. However, this is mitigated by the requirement for a strict markdown template, the instruction to redact sensitive data, and a mandatory "To resume" paragraph that warns future agents to verify all unverified claims against primary sources.

hand-off