Auto Context Manager

What This Skill Does

The Auto Context Manager (ACM) is an advanced project-aware engine designed to streamline AI interactions by automatically identifying and maintaining context across multiple active projects. Instead of requiring users to manually define the scope of every task, ACM parses user input to detect project-specific keywords, effectively mapping queries to the correct workspace. It acts as a bridge between the agent and your local file system, ensuring that project-specific memory, documentation, and skill sets are prioritized based on the active session context. By leveraging a local configuration file, the tool provides a persistent, privacy-focused way to keep your AI interactions organized and highly relevant.

Installation

To integrate this skill into your OpenClaw environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/imraxy/auto-context-manager

Once installed, the system will automatically initialize the ~/.auto-context/ directory. If the directory is missing, the skill will create the default projects.json file on the first execution. No external dependencies or API keys are required for operation.

Use Cases

• Multi-Project Development: Easily switch between different coding repositories without needing to restate project requirements to the AI.
• Administrative Organization: Manage separate contexts for work, personal side-projects, and academic research in a single terminal environment.
• Automated Workflow Routing: When a user issues a command, the agent can trigger specific utility scripts based on the project-ID identified by ACM.

Example Prompts

1. "How does the current module fit into the authentication flow of the web-app project?"
2. "Switch my context to the data-analysis project and list the active sprint tasks."
3. "Summarize the recent changes made to the configuration files in the client-dashboard project."

Tips & Limitations

• Project Keywords: To improve accuracy, add unique keywords to your projects.json. Distinctive terminology significantly raises the confidence score during detection.
• Local Data Only: Since ACM operates strictly locally, it cannot sync context between machines unless you manually manage the ~/.auto-context/ folder via version control.
• Confidence Thresholds: Always monitor the confidence score if the agent seems to suggest a mismatch; you can force a project switch manually using python acm.py switch <project_id> if the automatic detection is ambiguous.