run-research-and-save-files

Pass

Audited by Gen Agent Trust Hub on May 20, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides a shell script (scripts/init-corpus.sh) for initializing the research environment. This script is used to create directories and boilerplate files. It includes a security check (is_slug) that validates input arguments against a strict kebab-case regular expression, preventing potential command injection via user-controlled strings. The file creation logic uses printf with literal strings, ensuring that content is not interpreted by the shell.
  • [PROMPT_INJECTION]: The skill manages a multi-agent workflow that ingests untrusted data from web sources, constituting an indirect prompt injection surface. This is the primary function of the tool. The skill incorporates several architectural safeguards: subagents are isolated via disjoint filesystem write scopes, read scopes are restricted to prevent context leakage between agents, and the final synthesis is performed by a central orchestrator that reviews all collected evidence. The skill also enforces strict citation and verification rules to identify contradictions or missing data.
Audit Metadata
Risk Level
SAFE
Analyzed
May 20, 2026, 10:43 PM
Security Audit — agent-trust-hub — run-research-and-save-files