i2r-skill-quality-mode
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a meta-authoring mode for testing and iterating on other skills using a local evaluation pipeline.
- [SAFE]: The workflow involves local script execution (i2r.py) and processing local test files, with no evidence of remote code execution or data exfiltration.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided idea sketches during evaluation to detect failures. This ingestion of untrusted data is a known surface, but the skill implements a multi-stage validation pipeline (schema tests, gate checks) to verify outputs. 1. Ingestion points: evals/pressure-scenarios/ files. 2. Boundary markers: Absent. 3. Capability inventory: i2r.py subprocess calls and file management. 4. Sanitization: Stage 4 schema validation and Stage 5 gate checks.
Audit Metadata