The Agent Skills Directory

[COMMAND_EXECUTION]: Executes diagnostic commands via SSH on a local network device (raspberrypi.local). This includes managing Docker containers (docker compose), checking system memory (free), and monitoring disk space (df).
[EXTERNAL_DOWNLOADS]: Performs network requests to various API endpoints on the local host to retrieve trading status, current alerts, and market quotes.
[REMOTE_CODE_EXECUTION]: Automated scanner alerts regarding piped execution to Python are identified as false positives. The command curl ... | python3 -m json.tool utilizes Python's standard library for JSON formatting and does not execute external content as code.
[PROMPT_INJECTION]: The skill reads and processes external data from Docker logs and bot APIs which could serve as a vector for indirect prompt injection if those logs are influenced by malicious external signals. However, the localized scope significantly mitigates this risk.
Ingestion points: Remote Docker logs and bot API responses fetched from raspberrypi.local (SKILL.md).
Boundary markers: None observed in the diagnostic sequence.
Capability inventory: SSH shell execution and HTTP networking via curl.
Sanitization: None; raw log output is interpreted directly by the agent to create a narrative summary.

troubleshoot