specstory-link-trail
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes historical session logs that include untrusted content fetched from external websites during previous sessions.
- Ingestion points: The
parse_webfetch.pyscript reads and extracts content from markdown files located in.specstory/history/. - Boundary markers: The parsing logic specifically isolates content between XML-like tags such as
<tool-use>and</tool-use>to process tool outputs. - Capability inventory: The skill uses
Bashto execute local Python scripts andReadfor file access; the Python logic itself is limited to regex extraction and formatting without network or write access. - Sanitization: The
generate_report.pyscript escapes pipe characters in extracted summaries to prevent markdown table injection, andextract_urls_context.pyimplements domain-based filtering to exclude common system and noise URLs.
Audit Metadata