The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill utilizes a pattern where JSON data is fetched from a local Temporal REST API (localhost:8081) and piped directly into a Python interpreter for processing. While this structure matches remote code execution patterns, the script being executed is provided in plain text within the skill, and the target is the local loopback address, posing no risk of external code injection.
[COMMAND_EXECUTION]: Provides instructions to interact with the local container runtime using docker ps and docker exec. These commands are used to view logs and processes inside ephemeral task sandboxes for debugging purposes.
[COMMAND_EXECUTION]: Includes management commands (e.g., python manage.py cleanup_signals) and direct database mutations via curl to a local ClickHouse instance (localhost:8123) to manage test data state.
[DATA_EXFILTRATION]: The skill accesses local environment files (.env) using grep to verify configuration settings like CLICKHOUSE_DATABASE. It also reads sandbox agent logs from local object storage via a Django shell. No data is transmitted to external or third-party domains; all diagnostic information remains on the user's local system.

debugging-signals-pipeline