The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection by ingesting data from external, potentially untrusted sources such as Jira ticket descriptions, acceptance criteria, and comments.
Ingestion points: The skill uses bun run jira:sync-issues to fetch Story metadata, business rules, and acceptance criteria. It also uses bun run api:sync to fetch external OpenAPI/Swagger specifications.
Boundary markers: The skill instructs the agent to read specific synced markdown files rather than direct tool output to maintain context, and utilizes a structured spec.md format for planning.
Capability inventory: The skill can execute authored tests via bun run test, perform type checking, and run linting tools. Subagents have write access to the project's epic-level test-specs/ directory and session state files.
Sanitization: A mandatory human approval gate is implemented at the end of Phase 1 (Plan) before any code implementation (Phase 2) begins.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of OpenAPI/Swagger specifications from backend URLs configured by the user to synchronize TypeScript types.
Evidence: The bun run api:sync command downloads remote JSON specifications to api/openapi.json for type generation.
[COMMAND_EXECUTION]: The skill executes local developer tools and test runners to verify automation code quality and functionality.
Evidence: Standard execution of bun run test, bun run types:check, and bun run lint:check is performed during the Review phase.
[CREDENTIALS_SAFE]: The skill follows security best practices by requiring sensitive credentials (API tokens, passwords) to be stored in a .env file which is excluded from version control, rather than hardcoding them in scripts or instructions.

test-automation