The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches market and sentiment data from various external publishers including Alpaca, SEC Filings (via SerenDB), Google Trends, and Perplexity/Exa.
[COMMAND_EXECUTION]: Provides detailed instructions for setting up a local trigger server (run_agent_server.py) and configuring cron jobs on the Seren platform. These operations involve shell commands and Python script execution.
[DATA_EXFILTRATION]: Persists trading strategy results, PnL tracking data, and self-learning artifacts to a remote SerenDB instance. This is a core part of the skill's documented functionality for cross-session persistence.
[CREDENTIALS_UNSAFE]: The skill manages a SEREN_API_KEY used for authenticating with the Seren platform and Alpaca. It provides instructions for secure management via environment variables and .env files, and includes a registration step via an official API endpoint.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted external data from SEC filings and news summaries. However, it processes this data using deterministic keyword scoring and numerical clamping in strategy_engine.py rather than passing raw text back to the agent for instruction following.
Ingestion points: strategy_engine.py (SEC filings, Google Trends, News/Perplexity).
Boundary markers: Uses structured SQL queries and JSON responses for data retrieval.
Capability inventory: Network requests via requests and urllib, SQL execution via MCP and psycopg.
Sanitization: Numerical extraction and keyword counting with clamping are used to generate stock scores.

saas-short-trader