dev-api-design

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process API contract artifacts such as OpenAPI specifications, GraphQL SDL, and protobuf files. It lacks explicit instructions to sanitize these inputs or use boundary markers, creating a surface where malicious instructions embedded in a schema's description or metadata could influence the agent's behavior.
  • Ingestion points: API design workflows and contract reviews involving external schema files as described in SKILL.md.
  • Boundary markers: None present in the instructions.
  • Capability inventory: The skill templates involve executing package managers, CLI testing tools, and local scripts.
  • Sanitization: No input validation or sanitization is specified for the content of processed schemas.
  • [REMOTE_CODE_EXECUTION]: The skill references the execution of scripts and tools from sources that are not part of the provided file set.
  • Unverifiable Scripts: The SKILL.md and assets/express-nodejs/express-complete-api.md files instruct the agent to run scripts (append_learning.py and refresh-versions.py) located in external paths (agents-skills-feedback-loop/scripts/ and frameworks/shared-skills/scripts/). These scripts are not included in the skill's source files.
  • Piped Execution in Templates: The assets/oasdiff-ci.yml file contains a template for GitHub Actions that uses curl | sh to install the oasdiff tool from the tufin/oasdiff repository.
  • [EXTERNAL_DOWNLOADS]: The skill's templates and documentation describe downloading software from external sources.
  • Tool Installation: Fetches an installation script for the oasdiff utility from the tufin organization's public GitHub repository.
  • Dependencies: Standard development templates include commands to fetch packages from the NPM and PyPI registries.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 02:03 PM