structured-autonomy-implement
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions tell the agent to follow an external implementation plan, which provides an entry point for malicious instructions that could influence the agent's behavior.
- Ingestion points: The untrusted external implementation plan document ingested as input to the agent.
- Boundary markers: Absent. The prompt does not define specific boundaries or include instructions to ignore potentially conflicting commands within the plan content.
- Capability inventory: The skill grants the agent the capability to execute shell commands ("run the build or test commands specified in the plan") and perform file system writes ("implement ONLY what is specified in the implementation plan", "Update the plan document inline").
- Sanitization: Absent. There is no validation or filtering of the plan's contents before execution of commands or implementation of code.
Audit Metadata