The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (MEDIUM): Missing unverified local dependency.\n
The script parse_webfetch.py imports extract_url_from_context from extract_urls_context.py, but this file is not included in the skill package.\n
The absence of this file prevents verification of the extraction logic, which could potentially contain malicious operations or be leveraged for unauthorized file access given the Bash and Read capabilities.\n- Indirect Prompt Injection (LOW): Surface for instruction injection from processed web data (Category 8).\n
Ingestion points: parse_webfetch.py ingests data from .specstory/history/*.md, which contains raw content fetched from external websites via the WebFetch tool.\n
Boundary markers: Absent. The report generated by generate_report.py does not use delimiters or instructions to the agent to disregard embedded commands in the summarized content.\n
Capability inventory: The skill is granted Bash and Read tool permissions, which could be exploited if an attacker-controlled summary successfully injects instructions into the agent's context.\n
Sanitization: The skill performs basic character escaping for Markdown tables (replacing | with \|) but lacks sanitization against natural language instruction injection.

specstory-link-trail