composio-br

Pass

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill enables the agent to interact with over 1000 external applications (Gmail, Slack, GitHub, etc.), which involves reading data that may contain malicious instructions.
  • Ingestion points: Data retrieved via tools like GMAIL_FETCH_EMAILS, GITHUB_GET_AN_ISSUE, and NOTION_FETCH_BLOCK_CONTENTS.
  • Boundary markers: The instructions lack explicit requirements for the agent to use delimiters or ignore embedded instructions when processing external content.
  • Capability inventory: The skill utilizes shell commands (curl) and Python scripts (Playwright, requests) for automation and data retrieval.
  • Sanitization: No sanitization or escaping mechanisms for external data are described in the skill.
  • [COMMAND_EXECUTION]: The instructions direct the agent to use shell commands such as curl to perform tool searches, schema retrieval, and tool execution via the Composio gateway.
  • [DATA_EXFILTRATION]: The skill accesses sensitive credentials from environment variables (COMPOSIO_GATEWAY_AUTH, BROWSERBASE_API_KEY) and configuration files (/data/workspace/composio-gateway/.env). These are used to authenticate with the vendor-managed gateway (composio-gateway.fly.dev) and the Browserbase service.
  • [DYNAMIC_EXECUTION]: The skill suggests generating and executing local Python scripts to handle complex or recurring workflows, such as multi-step calendar operations or Twitter media uploads.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 4, 2026, 03:08 AM