convince-me
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data that could be controlled by an attacker, creating a vulnerability surface.
- Ingestion points: The skill reads external data including "original request, available spec or ticket, acceptance criteria, relevant changes, and existing tests" (SKILL.md, Section 1).
- Boundary markers: No specific delimiters or warnings to ignore embedded instructions are defined for the ingested data.
- Capability inventory: The skill executes shell commands, interactions, and scripts ("Run every proving command or interaction now") as defined in Section 3 of SKILL.md.
- Sanitization: No sanitization, escaping, or filtering of the external content is mentioned before it is used to design and execute proofs.
- [COMMAND_EXECUTION]: The skill's primary function involves identifying and executing shell commands and interactions within the agent's environment to verify work status.
- [DYNAMIC_EXECUTION]: The skill dynamically generates and executes "focused tests, scripts, fixtures, or disposable harnesses" at runtime to supplement existing verification methods.
Audit Metadata