gigio-execute-plan
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data from project files to direct the behavior of the agent and its delegated workers.
- Ingestion points: The skill reads plan files from the
.plans/directory and project-wide decisions fromPROJECT.mdat runtime to determine task ordering and execution scope. - Boundary markers: The instructions include a mandatory preamble for delegated workers to limit their capabilities (e.g., 'Do not spawn subagents', 'Do not invoke other skills'), but the processed input files themselves do not utilize explicit safety delimiters or escape sequences.
- Capability inventory: The skill uses
gitfor state verification, performs model handoffs (small-model-handoff), manages file checkouts (git-worktree-setup), and records results to disk. - Sanitization: It implements a 'Preflight' verification step (Step 1) that compares the plan against the actual state of the filesystem using
git diff, which serves as a integrity check for the ingested data.
Audit Metadata