The Agent Skills Directory

[PERSISTENCE]: The skill uses a prepare script in package.json to automatically configure git hooks by executing node scripts/wire-hooks.mjs, which redirects core.hooksPath to a local scripts/hooks directory. This ensures that the QA gate runs on every push attempt. While this is a common practice for development workflows and the skill provides instructions for disabling it, it remains a persistence mechanism that modifies local git configuration.
[COMMAND_EXECUTION]: The skill executes several local scripts and shell commands to perform its QA functions, including scripts/verify.mjs, scripts/wire-hooks.mjs, npm test, astro build, and check_seo.py. These commands are used to build the site, wire configuration, and run the testing suite.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data by reading rendered website content, including copy, metadata, alt texts, aria-labels, and JSON-LD (as implemented in placeholders.spec.ts and tone.spec.ts). This creates a vulnerability surface where malicious instructions embedded in the website being tested could potentially influence the agent's behavior during the code-fixing phase.
Ingestion points: Rendered website copy, <head> metadata, alt texts, aria-labels, and JSON-LD are parsed from the site pages.
Boundary markers: The instructions do not define specific delimiters or warnings to separate untrusted site content from the agent's operational instructions.
Capability inventory: The agent has capabilities to execute shell commands (npm test), modify source code files ("fix to green"), and perform git operations.
Sanitization: No specific sanitization, filtering, or escaping of the ingested website content is mentioned before it is processed by the agent.
[EXTERNAL_DOWNLOADS]: The skill mentions the requirement to install the Chromium browser binary using npx playwright install chromium. This is a standard dependency for the Playwright testing framework used by the skill.

website-qa