x9-idea-critic
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
claudecommand-line interface and system commands likemktempto handle temporary data securely.\n- [INDIRECT_PROMPT_INJECTION]: The skill evaluates user-provided ideas, which creates a potential surface for injection. It mitigates this via a 'sealed brief' system that isolates input and instructs sub-agents not to exceed the provided context.\n - Ingestion points: Captures user proposals and discussed outcomes from conversation history in SKILL.md.\n
- Boundary markers: Uses a sealed packet structure and explicit instructions to restrict sub-agent context in SKILL.md.\n
- Capability inventory: Invokes external CLI tools and manages sub-agent sessions as described in references/claude-cli.md and references/opencode.md.\n
- Sanitization: Prohibits the inclusion of secrets or credentials within the evaluation brief in SKILL.md.\n- [EXTERNAL_DOWNLOADS]: The skill references documentation for OpenCode agent setup at
https://opencode.ai/v2/docs/agentsin references/opencode.md.
Audit Metadata