The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run() in scripts/test_triage.py to execute its companion triage.py script. This is part of the skill's internal automated testing suite to verify the correctness of its attention-ranking logic and does not pose a security risk in this context.
[EXTERNAL_DOWNLOADS]: Fetches live financial data, corporate filings, and retail sentiment lists from well-known and legitimate services including Yahoo Finance (yfinance), SEC EDGAR (sec.gov), and Robinhood (robinhood.com). These operations are essential to the skill's core functionality.
[REMOTE_CODE_EXECUTION]: Executes the mermaid-cli utility via npx to render reasoning diagrams of the agent's decision-making process. This is a standard and widely-used tool for generating charts from text-based diagrams.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it analyzes external news articles and corporate filings which could theoretically contain adversarial text.
Ingestion points: scripts/smartmoney.py (SEC EDGAR filings), read-news (news content aggregator).
Boundary markers: External data is structured into JSON/Markdown blocks before injection, and subagent prompts are strictly delimited to enforce specific output formats.
Capability inventory: The agent can send notifications via the mkt skill, write research artifacts to local storage, and publish to the user's Notion workspace if configured.
Sanitization: Corporate filings are parsed using structured regex/XML logic. Unstructured news content is analyzed by AI agents using prescriptive persona instructions that require citation of specific facts, providing a degree of logical sanitization.

stocks-advisor