ralph-kage-bunshin-debug

Pass

Audited by Gen Agent Trust Hub on Mar 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes several shell commands to perform its diagnosis and reporting tasks. Evidence includes the use of grep to search code, agent-browser for UI inspection, and curl for reporting results to a local endpoint.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It reads content from potentially untrusted files like .ralph/workers/worker-N/PROGRESS.md, error messages, and source code. An attacker could inject instructions into these files to influence the agent's diagnosis or the data it sends externally via local commands.
  • [EXTERNAL_DOWNLOADS]: The skill recommends installing the @anthropic-ai/agent-browser package if it is not already present. This package is provided by a well-known, trusted organization.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 28, 2026, 02:34 PM