ralph-feedback-batch-protocol

Pass

Audited by Gen Agent Trust Hub on Mar 9, 2026

Risk Level: SAFEPROMPT_INJECTIONNO_CODE
Full Analysis
  • [PROMPT_INJECTION]: The skill establishes a workflow for processing human-provided feedback that influences the agent's internal state and replanning logic, introducing a surface for indirect prompt injection.
  • Ingestion points: Human feedback is recorded in feedbacks.md and processed by the orchestrator.
  • Boundary markers: The template uses YAML frontmatter and Markdown headers for structure, but lacks explicit instruction to ignore potential commands embedded within the feedback text.
  • Capability inventory: The skill modifies metadata.yaml to transition the orchestrator state to REPLANNING and passes control to a planner agent.
  • Sanitization: There are no mechanisms described for sanitizing or validating the feedback content.
  • [NO_CODE]: The skill is composed entirely of Markdown instructions and templates with no attached scripts, which eliminates the risk of direct malicious code execution within this specific component.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 9, 2026, 11:17 PM