finalize
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and follow project-specific checklists defined in local configuration files like
CLAUDE.mdorAGENTS.md. This ingestion of external data could lead to indirect prompt injection if those files contain malicious instructions meant to divert the agent. - Ingestion points:
CLAUDE.mdandAGENTS.md(specified inSKILL.mdsections 'How it works' and 'Project-Specific Checks'). - Boundary markers: Absent; the skill instructions direct the agent to work through found items directly.
- Capability inventory: Shell command execution (linting/testing); Git operations; memory updates; internal UI audit tool invocation.
- Sanitization: No specific content validation is defined within the skill itself.
- [SAFE]: The skill includes explicit instructions to scan changed files for security issues, specifically targeting hardcoded secrets and unsafe patterns in Section 3 (Code Quality Sweep).
Audit Metadata