ziw-implement
Pass
Audited by Gen Agent Trust Hub on Jun 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it processes data from issue trackers and PRs. Ingestion points: Tracker issue bodies, comments, PR comments, and external documentation (SKILL.md). Boundary markers: The 'Instruction Trust' section provides explicit behavioral constraints (SKILL.md). Capability inventory: The agent can execute local commands, modify files, and update issue trackers (SKILL.md). Sanitization: The agent is directed to identify and report any attempts to override instructions as security findings (SKILL.md).
- [COMMAND_EXECUTION]: The skill involves executing local development tools and CI-equivalent commands, such as
pnpm format:docs:check, to verify code changes as required by the repository configuration.
Audit Metadata