The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from numerous social media platforms and web archives, which could contain malicious instructions. Evidence chain: 1. Ingestion points: exports.py functions including scrape_page, sc_get, apify_run, and archive_fallback. 2. Boundary markers: The skill instructions do not mandate specific delimiters for the agent's output, although they do advise the agent to treat scraped Markdown as an 'extraction substrate' rather than absolute truth. 3. Capability inventory: The skill performs network requests via proxied_get and proxied_post and returns content for agent processing. 4. Sanitization: Instructions recommend that the agent validate extraction results and convert content into generic structured summaries.\n- [EXTERNAL_DOWNLOADS]: The skill fetches data from several external services including ScrapeCreators (api.scrapecreators.com), Firecrawl (api.firecrawl.dev), Apify (api.apify.com), and various web archive services (archive.org, archive.today). These network operations are the primary intended function of the skill and utilize a vendor-managed transparent proxy for authentication and billing.\n- [DYNAMIC_EXECUTION]: The helper script exports.py and the SKILL.md instructions involve modifying sys.path to include the skill's local directory. This is a standard and expected pattern in the agent environment to allow importing local modules like exports.py.

web-crawler