The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external and potentially untrusted data, specifically OpenAPI specifications and code diffs, which creates a surface for indirect prompt injection attacks where malicious instructions could be embedded in the specification.
Ingestion points: OpenAPI specification files (local or URL-based) and code diffs as described in SKILL.md and references/01-overview.md.
Boundary markers: The instructions lack explicit delimitation or boundary markers to distinguish untrusted specification data from the agent's internal reasoning or instructions.
Capability inventory: The workflow involves shell-based operations for code generation (openapi-ts) and CI validation scripts, creating a path for executed code to be influenced by processed specifications.
Sanitization: No automated sanitization or filtering of input content is specified before the agent processes the diffs.
[EXTERNAL_DOWNLOADS]: The skill includes instructions for downloading and installing development dependencies related to the OpenAPI generation toolchain.
Evidence: references/01-overview.md provides commands for installing @hey-api/openapi-ts, @tanstack/svelte-query, and zod via standard package managers.
[SAFE]: The skill provides proactive security guidance to prevent common vulnerabilities such as hardcoded credentials and improper TLS configurations.
Evidence: references/01-overview.md and references/05-ci-i-kontrola-kontraktu.md explicitly instruct the user to use environment variables for tokens, avoid hardcoding secrets in configuration files, and warn against disabling TLS verification (NODE_TLS_REJECT_UNAUTHORIZED=0) in production.

hey-api-openapi-review