feature-validator
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including git diffs, implementation files, and feature specifications, which could potentially contain malicious instructions designed to influence the agent's validation verdict. Ingestion points: The agent reads docs/specs/*.md, PROGRESS.md, feature_list.json, and implementation diffs (SKILL.md, Workflow Step 1 & 2). Boundary markers: No explicit boundary markers or instruction-isolation techniques are defined for processing the implementation diffs. Capability inventory: The skill can execute shell commands including init.sh, pnpm test, pnpm test:e2e, and manual smoke checks (SKILL.md, Workflow Step 3). Sanitization: There is no evidence of sanitization or filtering applied to the ingested implementation data before processing.
- [DYNAMIC_EXECUTION]: The skill is instructed to execute a shell script (init.sh) located within the repository it is validating. Evidence: In SKILL.md (Workflow Step 3), the agent is told to treat init.sh as a standard verification gate and execute it to run checks. Context: This behavior is intended for automated quality assurance and verification of local repository code, which is the primary purpose of the skill.
Audit Metadata