The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run in scripts/schedule.py and scripts/notify.py. In schedule.py, it is used to manage crontab entries for scheduling alerts. In notify.py, it executes notify-send for desktop notifications. Both uses are scoped to specific, non-malicious utility purposes and use shlex.quote or static binary paths to prevent shell injection.
[EXTERNAL_DOWNLOADS]: The skill fetches market data and earnings calendars from well-known financial services including Yahoo Finance, Alpaca, Nasdaq, Finnhub, and Alpha Vantage. These are legitimate data sources required for the skill's primary function and do not involve executing untrusted remote code.
[CREDENTIALS_UNSAFE]: The skill requires several API keys (e.g., ALPACA_API_KEY_ID, FINNHUB_API_KEY). However, it follows safe practices by retrieving these exclusively from environment variables or local .env files. No hardcoded secrets were detected in the source code.
[INDIRECT_PROMPT_INJECTION]: The skill processes external market data and news headlines which are interpolated into summaries. It implements a 'Neutral and factual' tone instruction to the agent to mitigate the risk of data influencing trade suggestions. Given the informational nature and the use of structured JSON for internal data passing, the surface for exploitable injection is low.
[PERSISTENCE]: The scripts/schedule.py script provides functionality to install a crontab entry for automated alerts. This is a documented and user-initiated feature of the skill to provide 'always-on' monitoring without an active agent session, rather than a hidden persistence mechanism.

watchlist-brief