alphagbm-smart-money
Warn
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill reads local transaction data from a user-specified file and transmits it to external hostnames for processing. While this is the primary function, it involves sending potentially sensitive data to the vendor's infrastructure.
- Evidence: scripts/run.py reads file content in read_transactions and sends it via fetch_json to hostnames defined in the HOSTS list.
- [DYNAMIC_EXECUTION]: The skill uses Python's dynamic loading capabilities to execute code from a local sibling file.
- Evidence: scripts/run.py uses importlib.util to load and execute scripts/review_engine.py at runtime.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from both user-supplied transaction files and remote API responses, creating a surface for indirect instructions.
- Ingestion points: scripts/run.py (via read_transactions and fetch_json).
- Boundary markers: Present in references/access.md, instructing the agent to ignore embedded commands.
- Capability inventory: scripts/run.py performs file reads and network POST requests.
- Sanitization: Input is validated via argparse and regex patterns.
Audit Metadata