agent-recovery

Pass

Audited by Gen Agent Trust Hub on Nov 18, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: The skill contains only instructional text and does not include any executable scripts, binaries, or automated command sequences.
  • [INDIRECT_PROMPT_INJECTION]: The recovery protocol involves reading untrusted data from GitHub Issues and Pull Requests. However, the skill explicitly mandates the separation of "structured records" from "reader-facing prose" and defines a robust "fencing identity" (execution generation) system to prevent unauthorized or concurrent state mutations. This architecture demonstrates an awareness of indirect injection surfaces and implements appropriate defensive boundaries. (File: SKILL.md; Ingestion: GitHub API; Capability: Git and GitHub mutations; Mitigation: Structured data schema and fencing tokens).
  • [SAFE]: The skill encourages security best practices, such as excluding secrets and absolute file paths from recovery checkpoints, and using idempotency keys for high-risk external side effects like database migrations or deployments.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 18, 2026, 12:11 AM