autonomous-loops
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes several autonomous loop patterns (such as the Infinite Agentic Loop and Ralphinho RFC-Driven DAG) that ingest external, potentially untrusted data to drive the agent's actions.
- Ingestion points: The patterns ingest data from specification files (
spec_file), RFC documents, and PRD documents found in the filesystem (e.g., indocs/orspecs/). - Boundary markers: The reference implementations do not explicitly describe the use of boundary markers, XML tags, or "ignore embedded instructions" warnings when interpolating these specifications into the agent's prompt.
- Capability inventory: These loops are granted extensive capabilities, including reading/writing files, executing arbitrary shell commands for builds and tests, and performing Git operations (branching, committing, pushing, and merging PRs).
- Sanitization: There is no mention of sanitizing, filtering, or validating the content of the ingested specification files before they are processed by the LLM, allowing a malicious specification to potentially highjack the autonomous loop's objectives.
Audit Metadata