auto
Pass
Audited by Gen Agent Trust Hub on Jul 31, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill serves as a legitimate intent router for a development toolset. It contains no executable code or scripts, relying entirely on structured prompts and existing platform tools.
- [COMMAND_EXECUTION]: While the skill invokes other skills via the 'Skill' tool, it implements a mandatory human-in-the-loop confirmation step ('Confirm') and a chain-of-thought reasoning step ('Classify') to prevent unintended or silent command execution. It explicitly forbids silent premium model upgrades and requires user consent for 'Heavy' tasks.
- [DATA_EXFILTRATION]: No network activity or access to sensitive credentials, SSH keys, or environment secrets was identified. The skill operates within the scope of standard development environment tools (Read, Grep, Glob).
- [PROMPT_INJECTION]: The skill does not contain instructions to bypass safety filters, ignore system prompts, or extract internal instructions. It includes specific guardrails to prevent recursion and ensure that routed skills' own security constraints are never bypassed.
- [INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data (the user's 'plain-english goal'). However, it mitigates this risk by presenting the parsed output for user confirmation before the 'Skill' tool is invoked, providing a robust boundary against automated exploitation.
Audit Metadata