iterative-loop
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill utilizes the 'ruflo' package via 'npx --no-install'. While the flag prevents new installations if not present, it still introduces a dependency on a non-standard external tool to manage hooks and memory states.
- [COMMAND_EXECUTION]: The skill's primary functionality involves executing shell commands for verification and orchestration, including 'npm test', 'eslint', 'tsc', and 'npx'. These are used within autonomous loops to determine completion.
- [INDIRECT_PROMPT_INJECTION]: The skill implements a self-correcting development cycle where the agent's next steps are guided by the output of external tools and previously written code.
- Ingestion points: Test results from shell commands, git history, and file contents processed during iterations (SKILL.md).
- Boundary markers: The skill suggests using explicit success criteria and 'promise' tags as boundaries, which help but do not fully mitigate adversarial input in processed files.
- Capability inventory: Shell execution capabilities for testing, linting, and state management ('npm', 'npx', 'tsc', 'eslint').
- Sanitization: The instructions do not specify sanitization or escaping of tool outputs before they are reintroduced into the prompt for the next iteration step.
Audit Metadata