quality-validation
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses coercive framing and behavioral nudges, such as referencing "failure memories" and stating "If you lie, you'll be replaced," to enforce strict adherence to its verification rules. This is designed to override the agent's default conversational and task-reporting patterns by creating artificial consequences for non-compliance.
- [INDIRECT_PROMPT_INJECTION]: The skill creates a loop where the agent must execute external tools (linters, test suites, build commands) and rely on their output to determine task success. This presents a vulnerability where a malicious payload in a file being tested could generate specific output to mislead the agent into believing a task is complete or successful when it is not.
- Ingestion points: The agent is instructed to read the "Full output" of verification commands (tests, builds, linters) as specified in SKILL.md.
- Boundary markers: No boundary markers or instructions to treat tool output as untrusted data are present in SKILL.md.
- Capability inventory: The instructions require the agent to execute various shell commands for testing and validation purposes.
- Sanitization: There is no logic provided to sanitize or validate the output from these external tools before the agent uses it to make completion claims.
Audit Metadata