Skip to content
Agent Memory
Skill

/start-jumbo-goal

Use when starting a Jumbo goal. Loads goal context, scope, and architectural constraints, then guides the agent through disciplined execution within defined boundaries.

From plugin
jumbocli
26518 skills
Install
$ npx -y skills add jumbocontext/jumbo.cli --skill start-jumbo-goal --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/start-jumbo-goal

Context preview

The summary Claude sees to decide when to auto-load this skill.

Use when starting a Jumbo goal. Loads goal context, scope, and architectural constraints, then guides the agent through disciplined execution within defined boundaries.

SKILL.md

start-jumbo-goal.SKILL.md
name: start-jumbo-goal
description: Use when starting a Jumbo goal. Loads goal context, scope, and architectural constraints, then guides the agent through disciplined execution within defined boundaries.

Start Jumbo Goal

**Prompt:** Start a Jumbo goal by loading its full context — objective, criteria, scope, architecture, components, decisions, invariants, and guidelines — then execute within the defined boundaries.

Why Structured Implementation Matters

When `jumbo goal start` is run, it assembles the goal's registered relations into structured implementation instructions. The objective defines what to build. The criteria define how success is measured. The scope defines boundaries. The architectural context (components, decisions, invariants, guidelines) defines constraints. Deviating from these instructions produces work that fails QA review.

Protocol

1. Start the Goal

jumbo goal start --id <goal-id>

This loads the full goal context. Read and internalize every section of the output before writing any code.

2. Understand the Implementation Instructions

The goal start output contains structured sections. Each section carries binding instructions:

Objective

The single purpose of this goal. All work must serve this objective.

Success Criteria

The specific, measurable conditions that determine whether the objective is met. Implementation is not complete until every criterion is satisfied.

Current Progress (if resuming)

If the goal was previously started, review what has already been accomplished and continue from there. Do not redo completed work.

Scope

  • **In Scope**: Files and areas you MUST work within.
  • **Out of Scope**: Files and areas you MUST NOT touch.

Solution Architecture

  • **Organization Style**: Namespaces (directory structures) and file names you introduce MUST maintain this style.
  • **Design Patterns**: You MUST leverage these patterns where applicable. If the goal requires a pattern not listed, register it with `jumbo architecture update`.
  • **Principles**: Artifacts you create MUST directly reflect these principles.

Relevant Components

Components that will be modified, created, or depended upon. Consider these while implementing.

Relevant Dependencies

External libraries involved in the implementation. Consider version constraints and contracts.

Relevant Decisions

Previous architectural decisions that inform or constrain this change. The solution must remain consistent with these decisions.

Invariants

Non-negotiable constraints. You MUST adhere to ALL invariants while implementing.

Guidelines

Coding standards, testing requirements, and process practices. You SHOULD follow these while implementing.

3. Execute Within Scope

  • Work only within the defined scope boundaries.
  • Follow all architectural constraints, invariants, and guidelines.
  • Track progress by documenting completed sub-tasks:
jumbo goal update-progress --id <goal-id> --task-description "<description>"

4. Maintain Context as You Implement

Register decisions, components, and corrections as they happen — not as a cleanup step after implementation.

  • **Architectural decision made** (e.g., chose pattern X over Y):
  jumbo decision add --title "Chose X over Y" --rationale "Because Z" --context "Background"
  • **New component created** (e.g., introduced a new service or builder):
  jumbo component add --name "ComponentName" --description "What it does"
  jumbo relation add --from-type goal --from-id <goal-id> --to-type component --to-id <component-id> --type involves --description "Created during implementation"
  • **User corrects your approach** (e.g., "don't do X, do Y instead"):
  jumbo invariant add --category "architecture" --description "Never do X because Y"
  jumbo guideline add --category "codingStyle" --description "Prefer X over Y"

Context registered now is served to future sessions. What you skip is lost.

5. Submit for Review

When all success criteria are met and implementation is complete:

jumbo goal submit --id <goal-id>

Rules

1. **Never deviate from the objective.** All code changes must serve the stated objective. 2. **Never violate scope boundaries.** In-scope files are your workspace. Out-of-scope files are off limits. 3. **Never violate invariants.** Invariants are non-negotiable constraints — no exceptions. 4. **Always track progress.** Use `jumbo goal update-progress` to document completed sub-tasks so future agents can resume if needed. 5. **Always submit when done.** Do not close or codify the goal yourself. Submit it for QA review. 6. **Do not enter plan mode for refined goals.** The goal context IS the plan. Execute immediately.

Read more
Ships withjumbocli

Memory and Context Orchestration for Coding Agents

Get the whole plugin