workflow-greenfield-init
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill contains instructions for the agent to execute a local Node.js script for maintenance of its internal contracts and documentation.
- Evidence:
node .claude/scripts/lib/workflow-skill-contract.cjs --writein SKILL.md. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted product ideas from users and perform deep-dive web research. This creates a surface for malicious instructions to enter the agent's context through external data.
- Ingestion points: User prompts are passed as context to
$start-workflow; external data is ingested via the$web-researchsub-skill. - Boundary markers: No explicit delimiters or isolation instructions are provided for user-provided context in this file.
- Capability inventory: The skill has permission to spawn sub-agents and orchestrate a wide array of tools for file manipulation, research, and planning.
- Sanitization: No sanitization or filtering logic is defined within this orchestration layer.
- [DYNAMIC_EXECUTION]: The skill dynamically loads and executes other functional skills from the local file system based on a registry file.
- Evidence:
Invoke $skill-name: load the registered .agents/skills/<name>/SKILL.md and execute it with Codex toolsin SKILL.md.
Audit Metadata