skills/clawic/skills/Chat Rooom/Gen Agent Trust Hub

Chat Rooom

Pass

Audited by Gen Agent Trust Hub on Apr 20, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes basic shell commands such as mkdir, cp, mv, and mktemp to manage its file-based coordination structure. These operations are limited to the local filesystem and are used for maintaining room logs and state persistence.
  • [PROMPT_INJECTION]: The coordination protocol requires the agent to read and respond to tasks, mentions, and summaries stored in markdown files within the workspace (.chat-rooom/rooms/). This creates a vulnerability surface for indirect prompt injection, as malicious instructions placed in these shared files could influence the behavior of participating agents.
  • Ingestion points: Content is ingested from summary.md, channel logs, and agent-specific inboxes.
  • Boundary markers: The protocol uses Markdown headers and formatted message blocks to delimit instructions.
  • Capability inventory: The agent can perform filesystem operations including directory creation and file movement.
  • Sanitization: No evidence of input validation or content sanitization was found for the data read from coordination files.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 20, 2026, 10:44 AM
Security Audit — agent-trust-hub — Chat Rooom