The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing user-provided OpenAPI specifications. This creates an attack surface where a malicious specification could contain instructions designed to influence the agent's analysis. However, since the skill's capabilities are restricted to local file operations (reading specs, writing reports, and editing the spec file), the risk is minimal and inherent to the tool's intended purpose.
Ingestion points: OpenAPI YAML/JSON files provided by the user for assessment.
Boundary markers: The skill instructions prompt the agent to strictly follow the provided WSO2 and Agent Readiness guidelines during analysis.
Capability inventory: File system access (Read/Write via fs and Edit tool) and command execution (spawnSync for Spectral).
Sanitization: The agent is instructed to perform a structural walk-through of the spec based on specific rules, which limits the impact of embedded prose instructions.
[COMMAND_EXECUTION]: The skill uses a local Node.js script (scripts/assess.js) to coordinate the assessment process. This script executes the spectral CLI tool to perform linting and uses standard system commands (open, start, xdg-open) to display the resulting HTML report in a browser. These are legitimate operations for an API development tool.
[EXTERNAL_DOWNLOADS]: The skill relies on the @stoplight/spectral-cli package, which is a well-known and trusted tool in the API development ecosystem. The skill instructions properly guide the user to install this dependency via NPM.

api-design