pragmatic-code-review
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill incorporates clear defensive instructions that prohibit the agent from executing the source code under review and explicitly prevent the reading of files ignored by
.gitignore, which helps protect against the accidental exposure of secrets or credentials. - [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it ingests untrusted code from a repository for analysis without utilizing boundary markers to isolate data from instructions.
- Ingestion points: The skill resolves a review scope into a list of files that are then read directly and fully into the agent context for review (SKILL.md).
- Boundary markers: Absent. No specific delimiters or "ignore embedded instruction" warnings are provided to the agent to distinguish the code being reviewed from potential adversarial instructions within that code.
- Capability inventory: The agent is authorized to read filesystem paths within the repository and write review reports and audit traces to the
docs/reviews/directory. - Sanitization: Absent. External code content is interpolated into the agent's context for evaluation without escaping or specific validation against injection patterns.
Audit Metadata