Agent Pipeline

What This Skill Does

The Agent Pipeline skill provides a standardized, automated workflow for managing complex software development tasks within OpenClaw. By orchestrating a sequence of specialized sub-agents, it ensures that every code change undergoes a rigorous, multi-stage validation process. The pipeline follows a linear progression: Researcher (optional) → Coder → Reviewer → Security → Tester → Commit. This systematic approach minimizes human error, enforces security standards, and maintains high code quality throughout the development lifecycle. By utilizing the Board API and Planner system, the skill ensures that all progress is documented, traceable, and actionable for project managers.

Installation

To integrate this skill into your environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/curbob/agent-pipeline Ensure that your OpenClaw instance has the necessary permissions to spawn sub-agents and access the auth-profiles.json configuration file, which must be propagated to all sub-agents to maintain authentication context.

Use Cases

This skill is ideal for teams or developers working on:

• Large-scale refactoring tasks where multiple eyes are needed for code review.
• Security-sensitive feature implementations requiring automated vulnerability scanning.
• Complex feature development that requires research into new APIs or unfamiliar library patterns.
• Standardizing development workflows across a distributed agent team to ensure consistent commit quality.

Example Prompts

1. "Agent Pipeline: Implement a new OAuth2 authentication endpoint for the user service. Start by researching the latest security best practices for our current framework."
2. "Use the Agent Pipeline to refactor the database connector. Skip the researcher step and proceed directly to coding; make sure the security auditor checks for SQL injection."
3. "Run the Agent Pipeline for the dashboard UI update. Please ensure the tester verifies all routing links are operational before the final commit."

Tips & Limitations

• Branching: Always branch from master using the feature/{short-description} convention.
• Failure Handling: If an agent reports a 'FAIL' status, do not proceed. Debug the specific issue, resolve it, and restart the pipeline at the failed stage.
• Logging: Maintaining board logs is mandatory. If an agent fails to update the board, the chain of custody for that task is broken.
• Limitations: This skill relies on the presence and correct configuration of sub-agents. Ensure that you have adequate compute resources allocated for spawning multiple agents simultaneously.