meta-kid-project-planner

Pass

Audited by Gen Agent Trust Hub on Aug 7, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted user input and interpolates it into multiple LLM prompts across its workflow. This creates a surface for indirect prompt injection where a user could attempt to override the skill's logic or manipulate downstream tool calls.
  • Ingestion points: 'inputs.user_message' is used throughout 'SKILL.md' (e.g., in 'preferences', 'fact_ledger', and 'deliver_project_pack' steps).
  • Boundary markers: The skill lacks robust delimiters or explicit 'ignore embedded instructions' warnings for the model when handling user data.
  • Capability inventory: The skill utilizes 'multi-search-engine', 'deep-research', 'weather', 'pptx' (file generation), and 'memory_save' (persistent state writing).
  • Sanitization: While 'xml_escape' and 'truncate' are used, they are insufficient to prevent complex prompt injection attacks designed to subvert the agent's logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 7, 2026, 08:36 PM
Security Audit — agent-trust-hub — meta-kid-project-planner