Skills
Navigate
skills/full-stack-skills/pencil-skills/pencil-mcp-get-style-guide/Gen Agent Trust Hub

pencil-mcp-get-style-guide

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The instructions mandate a 'CRITICAL' intent check, requiring the agent to verify that the user explicitly mentioned 'Pencil' before calling the tool. This is a best-practice safety constraint to prevent autonomous or accidental tool usage.
  • [NO_CODE]: No external code dependencies, scripts, or binary assets are included in the skill. The functionality is restricted to defined MCP tool calls within a controlled environment.
  • [INDIRECT_PROMPT_INJECTION]: The skill defines a data ingestion surface where user-provided input is used for tool parameters.
  • Ingestion points: The id and tags parameters in SKILL.md allow user-controlled values to be passed to the get_style_guide tool.
  • Boundary markers: Present. The skill uses 'Intent Recognition' and 'CRITICAL PREREQUISITE' blocks to define operational boundaries for the agent.
  • Capability inventory: Includes the get_style_guide tool; instructions suggest output is used in subsequent batch_design or set_variables calls.
  • Sanitization: Absent. The skill does not specify procedures for sanitizing the retrieved style guide metadata before it is used in downstream tasks.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 08:20 AM