The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script, scripts/agent-readiness-audit.py, using the system's Python interpreter to perform its analysis. This is a standard operation for performing deterministic checks on provided files.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from external sources, such as website HTML, robots.txt files, and product feeds in various formats (CSV, TSV, XML, JSONL). This establishes an attack surface where malicious instructions embedded in these files could attempt to influence the agent's behavior. The skill's design mitigates this by using a deterministic Python script for the primary analysis and by instructing the agent to report the script's status per check without replacing output with judgment.
Ingestion points: External files supplied by the user or fetched from a URL, including robots.txt, HTML exports, and product feed exports.
Boundary markers: No specific boundary markers or delimiters for the ingested data are defined in the instruction text.
Capability inventory: The skill has the capability to execute a local Python script and write files to the ${CLAUDE_PLUGIN_DATA} directory.
Sanitization: The skill relies on the underlying script to validate and parse inputs according to strict schemas (e.g., JSON-LD validation, Merchant Center attribute specs).
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch website content for auditing when the user explicitly provides a URL and uses the --fetch flag. It also references and links to official documentation and bot guidelines from well-known services and trusted organizations, including Google, OpenAI, Anthropic, Apple, and Perplexity.

agent-readiness-audit