The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell command execution over SSH on a remote host to manage containers and system health.
Evidence: ssh pi@raspberrypi.local 'cd /home/pi/keltner-bot && docker compose ps' used for status checks in SKILL.md.
Evidence: ssh "$PI_SSH" "cd $DEPLOY_PATH && docker compose logs ..." in scripts/bot-logs.sh.
[REMOTE_CODE_EXECUTION]: The skill instructions and scripts pipe output from network API requests directly into a Python interpreter for formatting.
Evidence: curl -s "http://${BOT_HOST}/api/health" 2>/dev/null | python3 -m json.tool in scripts/bot-restart.sh.
Evidence: curl -s http://raspberrypi.local:8080/api/reconciliation | python3 -m json.tool recommended in SKILL.md.
[DATA_EXFILTRATION]: The skill accesses and reads sensitive local configuration files and databases that may contain trading credentials or API keys.
Evidence: Reading the bot's environment configuration file: ssh pi@raspberrypi.local "cat /home/pi/keltner-bot/.env ..." in SKILL.md.
Evidence: Direct SQLite database access and modification: ssh pi@raspberrypi.local "sqlite3 /home/pi/keltner-bot/data/bot_state.db ..." in SKILL.md.
[EXTERNAL_DOWNLOADS]: The skill pulls software updates for the trading bot from a remote container registry.
Evidence: docker compose pull targeting ghcr.io/lgbarn/keltner-bot:latest (vendor resource from the author).
[INDIRECT_PROMPT_INJECTION]: The skill processes external logs and API responses, which could potentially contain malicious instructions if the target system is compromised.
Ingestion points: Docker logs via SSH and API responses from the bot dashboard.
Capability inventory: Remote shell access (SSH), network requests, and database manipulation.
Sanitization: Uses standard JSON parsing for API interactions.

bot-ops