ln-650-persistence-performance-auditor
Pass
Audited by Gen Agent Trust Hub on May 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust coordinator-worker architecture governed by explicit contracts for state management, phase transitions, and cleanup verification.
- [SAFE]: External data ingestion through research tools (WebSearch, WebFetch) is mitigated by an Epistemic Protocol that requires source attribution, priority-based tool fallback, and an actionability gate to prevent hallucinations or adversarial instructions from affecting the audit result.
- [COMMAND_EXECUTION]: Orchestration of sub-auditor workers (ln-651 through ln-654) is performed via a controlled Skill tool call rather than arbitrary shell execution, maintaining a high degree of task isolation.
- [DATA_EXFILTRATION]: No unauthorized data transmission was identified; the skill writes its findings and machine-readable summaries to local, platform-standard directories (.hex-skills/runtime-artifacts/) for project-local storage.
Audit Metadata