black-box

What This Skill Does

The Black Box skill provides a high-reliability, persistent audit logging mechanism for OpenClaw AI agents. By utilizing TiDB Zero as a robust, distributed backend, it transforms transient agent activity logs into indestructible records stored outside the local execution environment. Unlike standard file-based logging which is lost during container termination or system crashes, Black Box streams reasoning chains, execution errors, and critical state changes to the cloud in real-time. It effectively acts as a 'Flight Data Recorder' for autonomous agents, ensuring that every significant decision can be reconstructed even if the local environment is completely destroyed.

Installation

To integrate Black Box into your OpenClaw environment, ensure you have network access to TiDB Cloud on port 4000. Install the package directly from the OpenClaw repository using the command: clawhub install openclaw/skills/skills/lilyjazz/black-box. Once installed, you can configure your database credentials by setting the TIDB_HOST, TIDB_USER, and TIDB_PASS environment variables. If these are omitted, the skill will utilize the auto-provisioning fallback, creating a temporary database instance and caching the DSN in your local home directory for persistence.

Use Cases

This skill is indispensable for agents performing high-stakes tasks. Use it for compliance auditing where you must prove decision-making processes for regulatory bodies. It is also essential for post-mortem debugging of autonomous systems, as it allows developers to query the exact sequence of events leading up to a runtime failure. Furthermore, it serves as a long-term memory buffer, letting you track performance metrics across multiple agent sessions or deployments.

Example Prompts

1. 'Black Box, log a critical error: Authentication failed for the third consecutive time while accessing the remote API.'
2. 'OpenClaw, read the last 50 logs from the Black Box to investigate why the document summarization task was interrupted.'
3. 'Log the reasoning chain for the last task to the Black Box so I can verify the logic used for this data analysis report.'

Tips & Limitations

For production deployments, always provide your own TiDB credentials rather than relying on auto-provisioning to avoid losing access to logs if the temporary database expires. Note that while this skill ensures persistence, excessive logging of verbose reasoning chains may incur higher database costs, so prioritize logging state transitions and error events over every single sub-process output.