methodology-selector

Pass

Audited by Gen Agent Trust Hub on Jul 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted user data and is instructed to copy it into an output contract used by downstream processes.
  • Ingestion points: The skill reads user-supplied fields including "Task", "Audience", "Context", "Output format", "Validation", and "Human review / stop" as specified in SKILL.md.
  • Boundary markers: The output is encapsulated in a structured YAML block, providing a clear schema-based boundary.
  • Capability inventory: The resulting methodology contract is intended to be consumed by a downstream "goal-engineering" process that may have broader execution capabilities.
  • Sanitization: No sanitization or filtering is performed; the instructions explicitly require the agent to "copy the six input fields faithfully."
  • [COMMAND_EXECUTION]: The SKILL.md file contains instructions for the agent to execute a local JavaScript file for fixture validation: node /absolute/path/to/methodology-selector/scripts/self-test.mjs. The script itself performs integrity checks on the skill's manifest and references.
  • [EXTERNAL_DOWNLOADS]: Reference files such as references/aida.md and references/decision-matrix.md contain URL references to an external domain: https://loopengineering.app/methodology-skill-generator/.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 30, 2026, 04:48 PM
Security Audit — agent-trust-hub — methodology-selector