dossier

Pass

Audited by Gen Agent Trust Hub on Jun 19, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill retrieves organizational, financial, and legal data from well-known public and government services, specifically the SEC EDGAR (sec.gov) and ProPublica Nonprofit Explorer (propublica.org).
  • [COMMAND_EXECUTION]: Local Python scripts (citation_tracker.py, disconfirming_evidence_balance.py, source_tier_classifier.py) are used to implement core logic for audit logging, query balancing, and source tiering. These scripts use only standard library components.
  • [DATA_EXFILTRATION]: Session metadata and citation logs are stored locally in the user's home directory (~/.dossier_sessions/). This persistence is necessary for maintaining state across multi-step research phases and does not involve transmitting sensitive user data to external servers.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the web via WebFetch. While it lacks explicit instruction-delimitation in the phase descriptions, the operational scope (generating a report) and the use of reputable primary sources significantly mitigate risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 19, 2026, 07:17 AM
Security Audit — agent-trust-hub — dossier