connect-ai-manage
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on executing local Node.js and Python scripts to perform administrative tasks. Specifically,
scripts/cdata_workspaces.pycan execute theclaude mcp addcommand to register toolkits as MCP servers.\n- [CREDENTIALS_UNSAFE]: The skill manages Auth0 and Personal Access Tokens, storing them in a local cache file (connect-auth.json). While it implements machine-bound encryption for the refresh token and sets file permissions, the access token remains in plaintext, and the encryption mechanism is described by the author as not being cryptographically secret.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes data from the CData Connect AI platform which could be used to influence agent behavior.\n - Ingestion points: Administrative data (connection names, driver details, jobs) retrieved from
cloud.cdata.comand processed by scripts likecdata_jobs.pyandcdata_workspaces.py.\n - Boundary markers: Absent; data is displayed or used without specific delimiters to distinguish it from instructions.\n
- Capability inventory: The skill can perform network requests to the admin API, write to local configuration files (in
connect-cli.mjs), and execute commands via the bundled scripts (cdata_workspaces.py).\n - Sanitization: Absent; the skill lacks explicit sanitization of external platform data against prompt injection before processing.
Audit Metadata