update-ard-port

Pass

Audited by Gen Agent Trust Hub on Nov 18, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill fetches the ARD specification, JSON schema, and a conformance testing script from the ards-project/ard-spec repository on GitHub. These downloads are a core part of the synchronization workflow.
  • [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to interpret logic from an external conformance script and port those changes into local JavaScript files. This ingestion of external logic represents a potential attack surface.
  • Ingestion points: The skill downloads content from raw.githubusercontent.com/ards-project/ard-spec/ (referenced in SKILL.md).
  • Boundary markers: No specific delimiters or instructions to ignore embedded instructions are provided for the downloaded files.
  • Capability inventory: The skill possesses the ability to execute shell commands via node and yarn, and it modifies local source code files such as third-party/ard/ard.js (referenced in SKILL.md).
  • Sanitization: The procedure relies on manual side-by-side review of differences between upstream and local versions to identify and mitigate any malicious or unexpected changes.
  • [DYNAMIC_EXECUTION]: The command yarn build-ard-schema executes a script that generates a precompiled JSON schema validator (schema-validator.js) using the Ajv library. This is a standard dynamic code generation technique used to optimize validation performance in Node.js applications.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 18, 2026, 10:51 PM