ralph-specum-requirements

Pass

Audited by Gen Agent Trust Hub on Aug 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No malicious patterns or security risks were identified. The skill follows best practices for AI agent coordination, such as delegating work and requiring manual approval steps.
  • [PROMPT_INJECTION]: The skill ingests data from local project files (research.md, .progress.md) to generate requirements, creating a surface for indirect prompt injection. This is evaluated as safe because it is the core functionality of the skill and is governed by user approval gates.
  • Ingestion points: Step 2 and 5 in SKILL.md read research context and progress files.
  • Boundary markers: Absent.
  • Capability inventory: The agent can read/write project files and delegate tasks to sub-agents.
  • Sanitization: Absent.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 20, 2026, 07:36 PM
Security Audit — agent-trust-hub — ralph-specum-requirements