The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the alpaca command-line tool to perform trading operations, account lookups, and environment checks. All operations are executed through the local shell environment.
[EXTERNAL_DOWNLOADS]: Recommends installing the official Alpaca CLI from the vendor's GitHub repository or via Homebrew. These are standard installation methods for the tool the skill is designed to use.
[INDIRECT_PROMPT_INJECTION]: The skill processes external trade signals, creating a surface for potential instruction injection.
Ingestion points: Signal sources identified at the start of the trading workflow (SKILL.md, Section 4).
Boundary markers: Includes a 'Strategy confirmation checklist' and requires explicit user 'yes' confirmation before order submission (SKILL.md, Step 18).
Capability inventory: Shell command execution via the alpaca CLI for order submission, cancellation, and account management.
Sanitization: Employs a workflow where the agent must summarize and confirm all parameters with the user before execution to prevent the unintended execution of instructions embedded in signals.
[PERSISTENCE_MECHANISMS]: Includes templates for setting up automated trading via cron jobs, systemd timers, and launchd plists. These are provided as guidance for users wishing to automate their trading strategies.
[DATA_EXPOSURE_AND_EXFILTRATION]: While the skill interacts with sensitive API credentials in the environment, it contains explicit instructions for the agent to redact these tokens and avoid writing them to logs or session summaries. It also dictates that all session data remains in the user's local workspace.

alpaca-trading-paper-trading-cli