The Agent Skills Directory

[PERSISTENCE]: The skill implements automated background monitoring by configuring system-level persistence mechanisms.
Evidence: SKILL.md and references/scheduler-setup.md describe the installation of crontab entries and macOS LaunchAgents to execute the scan.sh script on a recurring schedule.
[COMMAND_EXECUTION]: The skill relies on executing various shell commands and external CLI utilities to perform its core functions.
Evidence: The skill uses okx CLI for market data, jq for JSON processing, and curl for network communications, as seen in scripts/scan.sh and the SKILL.md preflight steps.
[DATA_EXFILTRATION]: The skill transmits financial opportunity data from the user's environment to external notification platforms.
Evidence: scripts/scan.sh uses curl to send data to the Telegram Bot API and user-configured Lark webhooks.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection by processing external API data.
Ingestion points: Financial project data fetched via the okx CLI in scripts/scan.sh.
Boundary markers: Results are rendered using Markdown tables and block delimiters in various templates (e.g., templates/flash-earn.md).
Capability inventory: The agent has the ability to execute shell commands (exec), read/write local files, and perform network requests via curl.
Sanitization: Data is structured using jq during processing, but there is no evidence of specific sanitization or filtering of string content (like project names) to prevent prompt injection when rendered for the AI agent.

earn-hunter