goal-loop

Pass

Audited by Gen Agent Trust Hub on Aug 22, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow where the agent ingests a goal and repository context to autonomously execute state-changing actions.
  • Ingestion points: The skill accepts a <goal> argument and reads the repo context to infer completion criteria (SKILL.md).
  • Boundary markers: There are no defined delimiters or instructions to treat the goal or repository content as untrusted data, increasing the risk of the agent following instructions embedded within those sources.
  • Capability inventory: The skill allows for arbitrary state-changing actions and verification steps, routing through any available existing skills or tools (SKILL.md).
  • Sanitization: The instructions do not include any validation or sanitization logic for the inputs processed during the loop.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 22, 2026, 05:49 PM
Security Audit — agent-trust-hub — goal-loop