github-review-report
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes GitHub Pull Request reviews and comments, which are untrusted external inputs. These inputs are interpolated into a prompt for AI analysis without robust sanitization or strong boundary markers.
- Ingestion points: The skill fetches review content from GitHub PRs using the gh CLI as implemented in generator.py.
- Boundary markers: The AI prompt template in SKILL.md uses markdown headers (e.g., 原始Review内容:) as delimiters, but these are insufficient to prevent a malicious input from breaking out of its context.
- Capability inventory: The agent has access to the gh CLI tool (which can read/write PRs and comments) and can write local Excel reports using pandas.
- Sanitization: There is no evidence of sanitization, filtering, or escaping of the review content before it is processed by the AI.
Audit Metadata