anthropic-skilljar-extractor
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a course extraction workflow using a local Python script to scrape and organize content from anthropic.skilljar.com. All operations are consistent with the skill's stated purpose.
- [EXTERNAL_DOWNLOADS]: The extraction script fetches HTML from the user-provided Skilljar URL and downloads associated diagrams from Skilljar's CloudFront CDN (d7juhi4i8fsw0.cloudfront.net). These network operations are restricted to the primary service provider's infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external webpages which represents an attack surface for indirect prompt injection. Malicious instructions hidden in course notes could theoretically influence the agent during the enrichment steps.
- Ingestion points: The fetch function in scripts/extract_course.py ingests external HTML content via urllib.
- Boundary markers: No explicit delimiters or warnings are used when interpolating extracted notes into markdown files.
- Capability inventory: The skill has file system write access (os.makedirs, open) and network download capabilities (urllib.request).
- Sanitization: The script employs targeted regular expressions to extract specific curriculum elements and notes, providing a level of structural validation.
Audit Metadata