The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to perform local shell operations including npm install, npm run build, and execution of CLI tools like npx generate-tests and npx validate-plan. These commands are essential for preparing and running the skill's internal validation and translation logic.
[EXTERNAL_DOWNLOADS]: To set up its operational environment, the skill instructs the agent to run npm install, which fetches standard dependencies (such as zod) from the official npm registry. This is a routine setup process for script-heavy AI agent skills.
[DYNAMIC_EXECUTION]: The workflow involves generating and executing transient Node.js scripts using bash heredocs (`node
<<'NODE'andcat << 'SCRIPT' | node`). These scripts are used to perform schema validation and target analysis on user-provided data within the agent's runtime environment, following pre-defined logic templates.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided acceptance criteria as its primary input.
Ingestion points: User-provided .feature and .md files are read and parsed by the agent (defined in agents/assessment-mode.md).
Boundary markers: Data is isolated through structural transformations and JSON serialization during the conversion process.
Capability inventory: The agent has the capability to write files to user-specified directories and execute shell commands for build processes (defined in agents/conversion-mode.md).
Sanitization: The skill's translation engine (scripts/translate-plan-to-tests.ts) utilizes JSON.stringify for value interpolation and safe regex construction to mitigate simple code injection risks in the generated test files.

accelint-ac-to-playwright