skills/vllnt/skills/manage-rules/Gen Agent Trust Hub

manage-rules

Pass

Audited by Gen Agent Trust Hub on Nov 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and manage user, project, and nested instructions, which represents a potential surface for indirect prompt injection where untrusted data could influence agent behavior.
  • Ingestion points: The workflow in SKILL.md (Step 1) involves discovering active instruction sources, scopes, activation, and imports from potentially untrusted project environments.
  • Boundary markers: The SKILL.md file includes a 'Boundaries' section that explicitly instructs the agent to separate runtime-observed, source-simulated, and unknown precedence, which helps distinguish between different levels of instruction authority.
  • Capability inventory: The skill has the capability to perform "authorized validated edits" and code changes as defined in both SKILL.md (Workflow Step 5) and the quality-validation.md protocol.
  • Sanitization: The skill relies on a detailed defensive protocol in references/vstack/protocols/quality-validation.md. This protocol includes requirements for self-assessment, independent review, critique of assumptions, and traceable evidence ('PASS' verdicts) before any modifications are applied.
  • [NO_CODE]: The skill consists entirely of Markdown-based instruction and protocol files. It does not contain any executable scripts (e.g., Python, JavaScript, Shell) or binary files, which significantly reduces the direct execution risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 5, 2026, 08:18 PM