do-skill-craft
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized data access were detected. The skill provides high-quality instructions for the development and auditing of other agent skills, adhering to best practices for information hierarchy and performance optimization. It implements advanced reliability features such as an anti-confabulation protocol to ensure accuracy during reviews.
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external URLs and repository content during its 'distill' mode. Ingestion points: references/distill.md (Step 1). Boundary markers: Explicit markers for external data ingestion are absent. Capability inventory: Web-reading tools for distillation and file-writing tools for review outputs. Sanitization: No specific filtering logic is prescribed for the ingested content. This surface is inherent to the skill's analytical purpose and is mitigated by the required synthesis process.
Audit Metadata