blueprint-deck

Warn

Audited by Gen Agent Trust Hub on Oct 28, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [DYNAMIC_EXECUTION]: The script tools/check.mjs uses the Node.js vm module (runInNewContext) to execute code segments extracted from the local HTML engine file (blueprint-deck.html). This is used to dynamically load validation logic and icon metadata during the checking process.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from untrusted external sources to generate diagram content.
  • Ingestion points: The skill reads Markdown files, code repositories, PDFs, and wiki exports specified in the user configuration. It also uses a web tool to fetch external URLs for content and branding extraction (SKILL.md Section 1).
  • Boundary markers: The skill generates structured JSON within an HTML template. Instructions include specific escaping rules (e.g., writing </ as <\/) to prevent external content from breaking out of the JSON data block.
  • Capability inventory: The skill can write files, execute shell commands, run local Node.js scripts, and initiate network requests via a web tool.
  • Sanitization: Documentation in reference/spec.md states that the engine strips scripts, event handlers, and javascript: links from text fields before rendering, though the effectiveness of this sanitization relies on the engine's internal logic.
  • [COMMAND_EXECUTION]: The skill instructs the agent to perform multiple shell-level operations:
  • File management using cp or Copy-Item to duplicate the engine template.
  • Execution of local utility scripts using node SKILL_DIR/tools/check.mjs and node SKILL_DIR/tools/insert.mjs.
  • Starting local HTTP servers using python3 -m http.server or npx serve to perform browser-based validation, which exposes the local file system on a network port (SKILL.md Section 5).
  • Executing update commands for the skill itself (e.g., claude plugin update, gh skill update) when requested by the user.
  • [EXTERNAL_DOWNLOADS]: The skill fetches version information from the GitHub API (api.github.com) to check for updates. It also fetches user-provided websites to extract logos and brand colors using a web search/fetch tool.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 28, 2026, 03:26 PM