The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and analyzing untrusted external content, which creates a surface for indirect prompt injection.
Ingestion points: scripts/check-links.mjs and scripts/inspect-metadata.mjs fetch and parse HTML content, sitemaps, and robots.txt files from external URLs or local build directories.
Boundary markers: The scripts do not implement explicit boundary markers or instructions to the agent to ignore embedded commands when processing the audit findings, though they do provide structured output.
Capability inventory: The skill possesses capabilities for network requests (fetch) and reading local files (readFileSync). It does not have write access to the filesystem.
Sanitization: The skill performs basic HTML entity decoding but does not sanitize content for potential adversarial prompt injection patterns.
[DYNAMIC_EXECUTION]: The skill dynamically resolves and loads the Playwright dependency at runtime.
Evidence: scripts/lib/browser.mjs uses import() and createRequire().resolve() to search for and load playwright or playwright-core from the project's working directory or global modules. While restricted to specific packages, this involves runtime module resolution and loading code from computed paths.
[COMMAND_EXECUTION]: The skill instructions direct the agent to execute Node.js scripts that interact with the network and local filesystem to perform their tasks.
Evidence: SKILL.md contains bash commands for running scripts/inspect-metadata.mjs and scripts/check-links.mjs. These scripts are the core components of the discoverability audit functionality.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to retrieve and audit external website content.
Evidence: scripts/lib/pages.mjs uses the standard fetch API to retrieve content from URLs provided as arguments to the audit scripts. This is the intended behavior for the tool's purpose of site auditing.

discoverability-all