output-critic
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill defines a markdown-based evaluation protocol. It is entirely instructional and does not include any code, scripts, or tool invocations that could pose a security risk. It includes defensive guardrails such as automatically rejecting output that contains hardcoded secrets.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze potentially untrusted content (output from other steps or users). However, it lacks any dangerous capabilities (e.g., shell access, network requests, or file writes) that could be triggered by instructions embedded in the analyzed data.
- Ingestion points: The skill processes output produced in previous steps of an agentic workflow for evaluation (SKILL.md).
- Boundary markers: The protocol uses structured Markdown headers for scoring but does not define explicit sanitization boundaries for the input text.
- Capability inventory: None. The skill does not invoke any shell commands or external tools.
- Sanitization: None. The skill relies on the AI's internal reasoning to evaluate content.
Audit Metadata