Skills
skills/coder/agent-browser/slack/Gen Agent Trust Hub

slack

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from data retrieved from the Slack workspace.
  • Ingestion points: The agent ingests external, untrusted content from Slack messages and search results using agent-browser get text and agent-browser snapshot in SKILL.md and references/slack-tasks.md.
  • Boundary markers: There are no boundary markers or instructions to ignore embedded commands within the ingested data.
  • Capability inventory: The skill utilizes the agent-browser tool to perform navigation, clicking, form filling, and taking screenshots across all referenced files.
  • Sanitization: No sanitization or validation of the content retrieved from Slack is performed before processing.
  • [COMMAND_EXECUTION]: The skill makes extensive use of the agent-browser CLI tool via bash commands to automate browser actions.
  • Evidence: Bash blocks in SKILL.md and references/slack-tasks.md demonstrate the use of agent-browser connect, agent-browser open, and agent-browser click to programmatically control a browser session.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 07:22 PM