alphagbm-grid-plan
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a Python runner (scripts/run.py) to execute financial modeling and data retrieval logic.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external API data and local files, creating a surface for indirect prompt injection. Ingestion points: The skill retrieves market data, articles, and reports from the AlphaGBM API, and reads transaction history or snapshots from local files (scripts/run.py). Boundary markers: Instructions in references/access.md explicitly direct the agent to treat all retrieved data as untrusted and to ignore any embedded instructions. Capability inventory: The skill performs network operations via urllib.request and reads files using Path.read_text. Sanitization: No programmatic sanitization or validation of the content for prompt injection patterns is present; safety relies on LLM guardrails and natural language instructions.
- [DYNAMIC_EXECUTION]: The runner script uses the importlib module to load the bundled scripts/review_engine.py script for comparing analysis results.
- [DATA_EXFILTRATION]: The skill makes network requests to the vendor's official domains (alphagbm.com, alphagbm.zeabur.app) to retrieve financial data and perform analysis, using an API key retrieved from the system environment.
Audit Metadata