consensus-code-merge-guard

What This Skill Does

consensus-code-merge-guard acts as an automated governance layer for your engineering workflow. Unlike traditional CI checks that only verify syntax or basic test coverage, this skill introduces a multi-persona weighted decision engine. It evaluates incoming pull requests by synthesizing signals from technical metadata (security flags, test results, reliability markers) against a set of established persona requirements.

Upon evaluation, it returns one of three concrete states: MERGE, BLOCK, or REVISE. Beyond merely returning a status, it generates auditable artifacts that are recorded back to your board state, ensuring that every merge decision is transparent, reproducible, and compliant with your project’s governance policies. By utilizing the consensus-guard-core architecture, this skill ensures that your merge logic remains consistent across the entire project ecosystem.

Installation

To add this skill to your OpenClaw environment, use the provided package identifier from the OpenClaw hub. Run the following command in your terminal:

clawhub install openclaw/skills/skills/kaicianflone/consensus-code-merge-guard

Ensure that your environment has node and tsx installed, as these are the primary runtime dependencies. The skill operates without requiring external network calls during the decision-making process, maintaining a secure, localized execution profile.

Use Cases

• Automated PR Triage: Use it in a fully autonomous pipeline to filter out low-quality code before it touches your staging environment.
• High-Risk Repository Governance: Apply strict persona-weighted checks on critical production repositories to ensure security standards are strictly met.
• Standardized Code Audit Trails: Generate persistent historical logs of why specific code changes were rejected or approved, making external security audits significantly easier.
• Collaborative Engineering Teams: Use the external_agent mode to integrate human or secondary-agent consensus into the merge flow, balancing machine precision with human expert intuition.

Example Prompts

1. "OpenClaw, run the consensus-code-merge-guard on PR #442 using the current production security persona set. Report the final decision."
2. "Perform a revision check on this PR. If it is blocked, generate an artifact explaining the missing reliability markers and provide a summary to the developer."
3. "Evaluate the pending merge request from the UI team using 'external_agent' mode, aggregating votes from the senior dev and security bot provided in the input file."

Tips & Limitations

• Tip: Always verify your persona sets before execution. The quality of the merge decision is only as good as the reliability signals you provide to the skill.
• Tip: Utilize idempotent decision lookup to avoid redundant compute cycles if the PR state has not changed.
• Limitation: The skill does not perform internal persona set generation; you must ensure a valid persona_set_id is defined within your board state for mode="persona".
• Limitation: While it enforces policy, it cannot fix code errors automatically; it will only signal REVISE for the developer to intervene.