duck-review
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data such as code diffs and pull request text, which is an inherent attack surface for indirect prompt injection.\n
- Ingestion points: The skill accepts diffs, PR descriptions, and code chunks as input (Workflow step 1 in SKILL.md).\n
- Boundary markers: No explicit delimiter markers or warnings to ignore embedded instructions are defined for the ingested code blocks.\n
- Capability inventory: The skill is strictly focused on text analysis and does not have access to shell execution, file writing, or network operations, as specified in its internal boundaries.\n
- Sanitization: No specific sanitization or filtering of the input content is described in the provided instructions.\n- [SAFE]: No malicious patterns, obfuscation, or unauthorized data access were detected. The skill uses local reference files (GUARDRAILS.md and review-comment-examples.md) for its logic and formatting. The author metadata is consistent with the provided vendor context.
Audit Metadata