skills.rest
nowonbun
nowonbunCOMMUNITY
Explore all skills by this author →
GitHub

ai-collaboration-governance

Enforce runtime controls for Claude MCP collaboration with timeouts and healthchecks.

Updated Aug 27, 2026
Search Tags:#claude#mcp#timeout#healthcheck#cross-review#request-size-control#fallback-flow
One-click install
npx skills add https://github.com/nowonbun/nowonbun-harness --skill ai-collaboration-governance
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ai-collaboration-governance
Source: https://github.com/nowonbun/nowonbun-harness/tree/main/codex-skills/runtime-management_ai-collaboration-governance
Command: npx skills add https://github.com/nowonbun/nowonbun-harness --skill ai-collaboration-governance

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Prevents unbounded or fragile Claude MCP collaborations by enforcing deterministic timeouts, request size controls, healthchecks, and explicit fallback flows so implementations and reviews do not hang or produce inconsistent results.

Core Features & Use Cases

  • Request Classification: Classifies requests as heavy or light based on prompt length, item count, file count, and MCP references.
  • Healthchecks & Timeouts: Runs responsiveness checks and enforces a 5-minute per-call timeout with cooldown and fallback logic for repeated failures.
  • Fallback & Handoff Flows: Defines split-flow handoffs, two-session fallback behavior, and structured inputs/outputs for investigation, implementation, and review stages.
  • Delegated Logging & Policy Resolution: Delegates review output schema and logging format to a separate cross-review protocol while mandating CLAUDE as the policy source for review profiles.
  • Use Case: Apply these rules when orchestrating Claude-based cross-reviews for code changes or document audits to ensure resilience and auditable handoffs.

Quick Start

Apply the governance rules to classify a Claude MCP request, run healthchecks for heavy requests, enforce request size limits, and follow the timeout fallback flow if needed.

Frequently Asked Questions about ai-collaboration-governance

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent Claude MCP timeouts during cross-review pipelines?▼

To prevent Claude MCP timeouts, you can enforce a 5-minute per-call timeout with cooldown and fallback logic. This governance Skill applies runtime controls to ensure cross-review pipelines do not hang during implementation or verification stages.

What is a heavy versus light request classification in Claude MCP collaboration?▼

Heavy versus light request classification splits Claude MCP prompts based on length, item count, file count, and MCP references. This classification determines when healthchecks and split-flow handoffs are required for safe cross-review orchestration.

How do I set up request size limits and healthchecks for Claude MCP?▼

You set up Claude MCP request size limits and healthchecks by classifying requests as heavy or light, then running responsiveness checks for heavy requests. This gating prevents unbounded requests and ensures resilient handoffs.

Does this approach support two-session fallback flows for Claude MCP?▼

Yes, this approach supports two-session fallback flows and split-flow handoffs for Claude MCP collaboration. It enforces structured inputs and outputs for investigation, implementation, and review stages to handle repeated failures.

Can I delegate review logging to an external cross-review protocol with Claude?▼

Yes, you can delegate review output schema and logging format to a separate cross-review protocol while mandating CLAUDE as the policy source. This ensures auditable handoffs and consistent policy resolution across sessions.

When should I not use uncoordinated Claude MCP handoffs?▼

You should not use uncoordinated Claude MCP handoffs when processing heavy requests with large file counts or MCP references. Without split-flow handoffs and timeout fallback, heavy requests risk hanging or producing inconsistent review results.

Related Skills

Explore More Software Engineering Skills→
outfitter-devoutfitter-dev
28

claude-code-configuration

Manage Claude configuration files for MCP servers and environment variables.

Official
Intermediate
Stack-and-FlowStack-and-Flow
23

components-auditor

Audit React/TypeScript/Tailwind/Radix components against the shared contract and compliance matrix.

Official
Advanced
johnlindquistjohnlindquist
24

spider

Crawls websites with HTTP fetch, link extraction, and HTML parsing for JSON output.

Community
Intermediate
thejefflarsonthejefflarson
20

model-dos

Enforce token limits, throttling, and timeouts on LLM API endpoints.

Community
Advanced