The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the application's source code, existing feature files, and support files to generate test scenarios and custom step definitions. If these files contain malicious instructions, they could influence the agent's behavior during the test authoring process.
Ingestion points: Application source code, neighboring .feature files, and project support files (referenced in SKILL.md and references/maintaining-tests.md).
Boundary markers: No specific delimiters or "ignore" instructions are defined for processing external code or test files.
Capability inventory: The skill has the capability to write to the local file system (creating .feature and .js files) and execute browser tests via the letsrunit_run tool.
Sanitization: No explicit sanitization or filtering of the ingested content is documented.
[DYNAMIC_EXECUTION]: The skill involves generating and saving JavaScript code for custom step definitions into the project's support directory. These files are subsequently loaded and executed by the test runner, as described in references/custom-steps.md. While this is the intended purpose of the skill, it creates a surface for dynamic code execution.
[COMMAND_EXECUTION]: The skill uses automated tools, specifically letsrunit_run, to execute the generated Gherkin scenarios and custom step definitions. This allows the agent to run code it has just written, which is a necessary but sensitive capability for browser-based testing.

letsrunit-writer