execute-plan-loop
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from external sources (the 'plan' and 'spec' provided by the user or project files) which directly guides the agent's actions during the loop. This content can influence the shell commands executed for tests, linting, and manual environment setups.
- Ingestion points: Reads the project 'plan' and 'spec' (SKILL.md), and reads state from
.scratchpad/<slug>/progress.md. - Boundary markers: None identified; instructions do not specify using delimiters or warnings when processing plan-based data.
- Capability inventory: Executes shell commands for tests, linting, and manual environment management (SKILL.md); writes to the file system (
progress.md,loop-log.md,manual-env.sh); and performs git operations. - Sanitization: No explicit sanitization or validation of the plan or spec content is mentioned before it is used to determine execution steps.
Audit Metadata