argo-wide-research

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill acts as an aggregator for untrusted external data retrieved via web research, which presents an attack surface for indirect prompt injection.
  • Ingestion points: Data enters the context during the 'worker retrieval' phase where the argo-researcher fetches content from external HTTP(S) URLs.
  • Boundary markers: The process mandates a structured return format including explicit fields for summaries, findings, and a source ledger, which helps separate external content from internal logic.
  • Capability inventory: The skill leverages sub-agents capable of executing Bash commands and reading local files.
  • Sanitization: The 'Quality Gate' mechanism requires source ledger validation and cross-language verification to prevent data poisoning from a single source from corrupting the final consensus.
  • [COMMAND_EXECUTION]: The skill uses Bash to invoke specific sub-agent capabilities.
  • The configuration specifies that argo-researcher is limited to a tool whitelist involving only atomic commands and file reading, which prevents recursive tasking or arbitrary command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 06:54 AM