weaver-plan

Author goal-focused Weaver Hub Manager-mode mission plans and land them via Brain or local YAML.

Updated May 12, 2026
One-click install
npx skills add https://github.com/contexthub-md/marketplace --skill weaver-plan-contexthub-md
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: weaver-plan
Source: https://github.com/contexthub-md/marketplace/tree/main/contexthub/skills/weaver-plan
Command: npx skills add https://github.com/contexthub-md/marketplace --skill weaver-plan-contexthub-md

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Writing mission plans that Weaver Hub can actually execute is error-prone: over-specified step scripts make the Manager-mode PM fake completion, schema fields like execution_target are silently ignored, and malformed dod_gate checks can orphan missions or make phases unpassable. This Skill encodes the verified schema rules and the proven phase patterns so plans land correctly the first time. ## Core Features & Use Cases - Goal-focused plan authoring: Structures phases as goal + acceptance criteria + guardrails instead of imperative step scripts, with explicit completion: goal_gated vs task_list declaration per phase. - Proven phase pattern: Applies the design -> HOLD-for-PASS human checkpoint -> implement -> batched deploy/dogfood structure, with per-todo model-tier discipline (Opus/Fable for design, Sonnet for execution, Haiku for mechanical sweeps). - Two landing paths: Writes local plans/*.yaml for fast orchestrator iteration, or lands plans through ch_create_mission_with_plan so they flow over MQTT through Brain, DynamoDB, and the retained-plan rehydration pipeline. - Use Case: Turn a roadmap item like "model-tier budget engine" into a Weaver-executable Manager-mode mission, verify it with ch_get_mission_plan, and repair fields in place with ch_edit_mission_plan. ## Quick Start Ask the agent to author a Weaver Manager-mode mission plan for your roadmap item and land it through Brain so Weaver picks it up over MQTT.

Frequently Asked Questions about weaver-plan

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

FAQPage Schema
How do I create a mission plan that Weaver Hub can execute?

Author phases as goal plus acceptance criteria plus guardrails rather than step scripts, then land the plan via ch_create_mission_with_plan so Brain writes it to DynamoDB and mqtt-bridge republishes it to Weaver. Alternatively write a local plans/*.yaml file and call start_mission for fast orchestrator-only iteration.

What is the difference between weaver-plan and mission-forge?

mission-forge is the generic idea-to-ContextHub-mission pipeline including AgentCore Mode 2 generation. weaver-plan is for Weaver Hub specifically, encoding goal-focused Manager-mode authoring, the design-checkpoint-implement pattern, and Weaver schema traps like the ignored execution_target field.

Why does my Weaver mission phase never complete?

A common cause is a file_exists or file_contains check in a dod_gate without a repo set on a multi-repo mission; the path resolves against the mission worktree root where only repo directories live, so the check can never pass. Always set repo on file checks when the mission owns more than one repo.

Can I assign a model per todo in a Weaver plan?

Not through a typed field yet. Model is chosen per execution mode in Weaver settings, so a per-todo model tier must ride in the todo metadata bag, which the MQTT PlanTodo wire shape carries. Assign tiers by altitude: Opus/Fable for design and review, Sonnet for execution, Haiku for mechanical sweeps.

When should I use goal_gated versus task_list completion?

Use goal_gated for outcome-shaped work where done must be proven by a machine-verifiable Definition-of-Done gate, such as a command with expected exit code. Use task_list for concrete known-shape work like docs phases or mechanical refactors where each todo completes on its own commit evidence.