The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external, potentially untrusted data such as PRDs, user stories, and bug reports to generate executable test code.
Ingestion points: Processes external inputs in Step 1 (Requirements Extraction) to drive the pipeline.
Boundary markers: Step 1 distinguishes between explicit and implicit requirements; Step 7 mandates human review of all generated outputs.
Capability inventory: Generates test code for various frameworks and utilizes verification tools including tsc, pyflakes, and grep for static analysis.
Sanitization: Implements a mechanical verification gate (Verification section) that uses linters and type-checkers to catch hallucinated APIs before the code is executed.
[COMMAND_EXECUTION]: Recommends the execution of various CLI tools for project initialization and code verification.
Evidence: Commands include npx playwright init-agents, npx tsc --noEmit, python -m pyflakes, and ruff check.
Evidence: Provides a shell script snippet using grep and sed to validate that generated test selectors exist within the target codebase.
[EXTERNAL_DOWNLOADS]: Mentions downloading official tools from well-known services.
Evidence: References npx @playwright/mcp@latest to fetch the Playwright MCP server.

ai-test-generation