quick-iteration
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains explicit instructions to override the agent's safety and verification protocols. The phrase "you just do it without running any checks" is a direct attempt to remove standard operational constraints and safety filters.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon data from external sources (specifically Pull Request branches) while explicitly disabling the agent's validation mechanisms. This creates a significant surface for adversarial content in a PR to be executed or followed without the usual safety guardrails.
- Ingestion points: Data provided via Pull Request (PR) branches and associated changes (SKILL.md).
- Boundary markers: The instructions provide no delimiters or markers to separate the agent's instructions from potentially malicious data in the PR.
- Capability inventory: During the iteration, the agent likely retains its standard toolset for file modification and command execution.
- Sanitization: There is a complete lack of sanitization, as the skill explicitly commands the agent to avoid running checks on the input it receives.
Audit Metadata