create-verification-skill
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze arbitrary content from a local repository to generate verification instructions. Maliciously crafted documentation, configuration files, or code comments in the target repository could potentially influence the agent's behavior during the generation or verification phases.
- Ingestion points: Local repository files (READMEs, package scripts, Makefiles, source code) accessed during the "Interview the repo" step in
SKILL.md. - Boundary markers: None identified in the provided instructions to delimit untrusted repo content from agent instructions.
- Capability inventory: Command execution via shell (e.g.,
node scripts/check-maps.mjs, app launch commands), file writing to.agents/skills/, and browser automation viacontrol-notes. - Sanitization: No explicit sanitization or validation of the repository's content is described before it is processed or interpolated into generated instructions.
- [COMMAND_EXECUTION]: The skill requires the execution of repository-specific build and start commands, as well as specific maintenance scripts like
node scripts/check-maps.mjs. This grants the agent the ability to execute arbitrary code defined within the target repository's environment.
Audit Metadata