The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.call and os.execv in run_acceptance.py to launch the behave test runner and gherkin-lint. These calls are localized to the project environment and are fundamental to the skill's purpose of providing an acceptance testing framework.
[EXTERNAL_DOWNLOADS]: The instructions and configuration files reference standard development tools and libraries, including behave, cucumber-js, cheerio, and gherkin-lint. These are well-known packages in the testing community and are used for their intended purposes.
[INDIRECT_PROMPT_INJECTION]: The skill implements an ingestion surface for Markdown-based Gherkin specifications, but incorporates strong structural boundaries. * Ingestion points: Specification files located at openspec/specs/**/spec.md and openspec/changes/**/spec.md (processed in extract_gherkin.py and extract-gherkin.cjs). * Boundary markers: The extraction logic uses explicit H1-H4 Markdown headings and column-0 ```gherkin fences to isolate executable test steps from descriptive prose. * Capability inventory: The extracted feature files are passed to local test runners (behave, cucumber-js) and linters (gherkin-lint) as part of the local development workflow. * Sanitization: The extraction process normalizes Gherkin keywords and strictly ignores all text outside of designated fences, effectively preventing non-test prose from being executed as Gherkin structure.

acceptance-test-authoring