logic-whiteboard

Warn

Audited by Gen Agent Trust Hub on Nov 12, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill executes code from a dynamically computed path at runtime.
  • Both scripts/doctor.mjs and scripts/whiteboard.mjs calculate an external engine path using either the ARTICLE2MEDIA_ENGINE environment variable or a relative path ../../article2media/engine/scripts/.
  • The scripts use await import() to dynamically load and execute code (doctor.mjs and render.mjs) from this computed location.
  • This mechanism allows the skill to execute code that is not contained within its own package, creating a dependency on external local state that could be manipulated.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data that is used to drive logical analysis and output generation.
  • Ingestion points: The skill processes article.md (which may contain arbitrary text from URLs or transcripts) as specified in SKILL.md and references/mode-whiteboard.md.
  • Boundary markers: There are no explicit instructions or markers used to isolate the untrusted article content from the agent's instruction context during processing.
  • Capability inventory: The skill has the ability to execute shell commands (via Node.js) and perform file system read/write operations.
  • Sanitization: No sanitization or validation of the input article's content is described, increasing the risk that malicious instructions embedded in the article could influence the agent's behavior during the '论证账本' (argument ledger) creation phase.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 12, 2026, 01:08 PM