nocobase-dsl-reconciler

Warn

Audited by Gen Agent Trust Hub on Apr 26, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill instructions (SKILL.md) and multiple scripts such as 'scripts/cleanup-copy.py' and 'scripts/copy-data.py' include hardcoded default credentials like 'admin@nocobase.com' and 'admin123' for development database and application access.
  • [DATA_EXFILTRATION]: The authentication logic in 'src/client/auth.ts' scans sensitive configuration files of other AI agent applications, specifically searching for Bearer tokens in '/.claude/settings.json', '/.claude/settings.local.json', and '~/.kimi/mcp.json'.
  • [COMMAND_EXECUTION]: The skill makes extensive use of the shell to execute system commands like 'git', 'npx', and 'psql', creating a potential attack surface for command injection if workspace configuration files are maliciously crafted.
  • [PROMPT_INJECTION]: The skill has a high surface area for indirect prompt injection. 1. Ingestion points: The skill reads and processes YAML, JavaScript, and SQL content from arbitrary project files in the 'workspaces/' directory. 2. Boundary markers: There are no delimiters or explicit instructions to ignore embedded commands when parsing these files. 3. Capability inventory: The skill possesses 'shell' access, performs network requests via 'axios', and executes raw database queries through 'psql'. 4. Sanitization: No sanitization or validation of the external file content is performed before it is used in system commands or deployed to the target NocoBase server.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 26, 2026, 12:26 PM