agent-loop
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Executes local CLI tools including
joelclaw,bun,kubectl, andslogto manage infrastructure and monitor agent runs. It specifically utilizescodexwith high-privilege flags (--sandbox danger-full-access,--ask-for-approval never) for autonomous code implementation. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates external project data into the agent's context.
- Ingestion points: The skill reads
prd.jsonandprogress.txtfrom the project root directory. - Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are documented for these inputs.
- Capability inventory: The pipeline has the ability to execute sandboxed commands with full access, commit code to git, and trigger Kubernetes rollouts.
- Sanitization: No sanitization or validation logic for the content of the PRD or progress files is mentioned.
Audit Metadata