make-verify
Warn
Audited by Gen Agent Trust Hub on Nov 2, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The skill is designed to programmatically generate other skills and helper scripts that execute shell commands. It directs the agent to create a
verifyskill (.agents/skills/verify/SKILL.md) and associated scripts that drive product interfaces and perform system checks, which are then executed locally during the verification phase.- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external data (the project repository) to identify entry points, commands, and expected outputs. It lacks a mandatory evidence chain or sanitization process to prevent malicious instructions hidden in the codebase from influencing the generated verification logic. - Ingestion points: Project source code and repository structure (SKILL.md Step 1).
- Boundary markers: Absent; the skill does not specify delimiters to separate untrusted code analysis from the generation of actionable commands.
- Capability inventory: The generated skill has the capability to execute shell commands, perform file system operations, and manage browser sessions (SKILL.md Step 2/5).
- Sanitization: Absent; the instructions do not require validation of command strings or handles extracted from the repository before they are included in the generated verification skill.
Audit Metadata