data_analysis
Fail
Audited by Gen Agent Trust Hub on Apr 2, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill provides a Python template containing a hardcoded 's3_access_key_id' ('rTFidVYcxQjiYCZOXhm8') and 's3_secret_access_key' ('BOEhYByB6e645RKiVTt2kvlMR2qCz2jVe90IfD5F') for a MinIO endpoint.\n- [COMMAND_EXECUTION]: The skill documentation instructs the agent to execute discovery commands using 'find' and run Python code using 'uv run'.\n- [REMOTE_CODE_EXECUTION]: The skill uses 'INSTALL httpfs;' within DuckDB to download and load a binary extension from an external repository at runtime.\n- [EXTERNAL_DOWNLOADS]: The skill configures connectivity to an external S3-compatible storage service ('minio:9000') and downloads necessary extensions for data handling.\n- [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection.\n
- Ingestion points: Processes data from local Parquet, CSV, and JSON files, and remote Delta tables in SKILL.md.\n
- Boundary markers: Absent; no instructions are provided to the agent to treat data as untrusted or to ignore instructions embedded in the data.\n
- Capability inventory: Ability to execute shell commands ('find') and Python scripts ('uv run') that can perform network and file operations.\n
- Sanitization: Absent; the skill does not define any validation or sanitization logic for the data being queried.
Recommendations
- AI detected serious security threats
Audit Metadata