create-verification-skill

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill automates the generation and execution of shell commands (such as Launch, Doctor, and Drive) tailored to the repository's environment. It specifically instructs the agent to execute these generated instructions to prove the skill's functionality.
  • [PROMPT_INJECTION]: The skill performs an "interview" of the repository's codebase and documentation to determine how the app should be driven, which creates a surface for indirect prompt injection.
  • Ingestion points: SKILL.md (Step 1) specifies reading the codebase, READMEs, and package scripts to determine app surface and drive methods.
  • Boundary markers: The skill does not define explicit delimiters or instructions to ignore embedded commands in the repository data.
  • Capability inventory: The agent is granted capabilities to write files (.cursor/skills/verify-<app>/SKILL.md), execute shell commands (Step 4), and interact with browsers or PTY sessions.
  • Sanitization: No specific sanitization or filtering of repository-provided content is mentioned before it is used to generate executable instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 09:58 AM
Security Audit — agent-trust-hub — create-verification-skill