handoff
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes session state from external handoff files, which presents a potential surface for instructions embedded in data.
- Ingestion points: Handoff Markdown files read during the 'Resume' process (SKILL.md, Resume Step 1).
- Boundary markers: Uses a strictly defined Markdown document contract and schema specified in REFERENCE.md.
- Capability inventory: Performs file system operations, git state management, and invocation of other installed skills (e.g., commit, make-pr) as defined in SKILL.md.
- Sanitization: Employs a multi-scan redaction process for secrets and performs comprehensive validation of paths, branches, and PR statuses before trusting resumed data.
- [DATA_EXPOSURE]: The skill includes an explicit 'Redaction' section in REFERENCE.md and a triple-scan requirement in SKILL.md (Create Step 4) to prevent the accidental leakage of credentials, tokens, private keys, and environment variables into handoff documents.
Audit Metadata