maintain-verification-skill
Pass
Audited by Gen Agent Trust Hub on Jul 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes project source files, feature maps, and documentation to identify "doc drift." This creates a surface for Indirect Prompt Injection, where malicious instructions embedded in the project source or feature files could influence the agent's behavior during the audit.
- Ingestion points: Project source code, feature files (e.g.,
features/*.md), and target verification skill files located in.agents/skills/or similar directories. - Boundary markers: None explicitly defined to separate untrusted source content from agent instructions.
- Capability inventory: File writing (restricted to the verification skill directory), command execution (launch/teardown flows for servers/CLIs), and spawning subagents for concurrent analysis.
- Sanitization: No explicit sanitization or instruction-ignoring markers are defined for the ingested source code.
- [COMMAND_EXECUTION]: The skill is designed to follow launch and teardown flows defined in other "target" verification skills. This involves executing shell commands to start servers, UIs, or CLIs for live verification passes. While restricted to project-local skills, this facilitates the execution of scripts and commands defined within the repository's configuration.
Audit Metadata