stockbee-setup-fluency-trainer
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The
FMPClientinscripts/build_model_book.pyperforms network requests tofinancialmodelingprep.comto fetch historical stock price data. This is a well-known financial data service used for the skill's primary purpose of calculating stock performance metrics. - [COMMAND_EXECUTION]: The skill requires the execution of a Python script (
scripts/build_model_book.py) with various subcommands (ingest,update,summarize) to process data and generate reports. These commands are intended to be run by the user or the agent to manage the local model book. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface as it processes external data to generate Markdown reports read by the agent.
- Ingestion points: Data is ingested from user-provided JSON files via the
--screener-jsonand--prices-jsonarguments inscripts/build_model_book.py. - Boundary markers: The skill does not use explicit boundary markers or 'ignore' instructions when interpolating ingested data (like symbols, tags, or notes) into the generated Markdown reports in
write_markdown_report. - Capability inventory: The skill has the capability to write to the file system (generating JSON and Markdown reports) and perform network requests (via the FMP API).
- Sanitization: Ingested strings undergo minimal sanitization beyond basic normalization (e.g., uppercase conversion for symbols), allowing potentially malicious strings embedded in screener reports to be rendered into reports the agent later processes.
Audit Metadata