Joint Planner — A doctrinal joint planning copilot for Claude Code

A Claude Code plugin that walks military planners through the seven-step Joint Planning Process (JPP) from Joint Publication 5-0, Joint Planning (Joint Chiefs of Staff, 1 December 2020). Provides structured plan workspaces, doctrinal critique, and artifact generation. UNCLASSIFIED only.

---

What this is

joint-planner is a Claude Code plugin that turns a planning session with Claude into a doctrinally faithful walk through the Joint Planning Process. It supports four modes:

• Sherpa — walk me step-by-step through a JPP step.
• Coach — explain doctrine, quiz me, walk a worked example.
• Critic — red-team my draft against the doctrinal rubric.
• Generator — produce an artifact (mission analysis brief, CONOPS, OPORD shell, comparison matrix, etc.).

The plugin is grounded in a Practitioner Field Guide to JP 5-0 that ships with the plugin (references/jp5-0-field-guide.md). Every skill embeds or references the relevant field-guide sections, and every plan artifact is validated against an explicit rubric drawn from the field guide.

What it is not

• Not a TPFDD/TPFDL generator. Force-flow data is out of scope.
• Not a CONPLAN/OPLAN annex generator. v1 supports plan Levels 1 (Commander's Estimate) and 2 (BPLAN). Annexes A–Z are deferred.
• Not classified-capable. This plugin runs through Anthropic's API and reads/writes files on your local disk. Do not enter classified information.
• Not a substitute for JP 5-0 itself. For authoritative doctrine, ROE/RUF questions, multinational ratification details, and exact format guidance for orders, defer to the source publications listed at the bottom of the field guide.

Hard rule

UNCLASSIFIED only. This plugin is for unclassified training and unclassified planning support. Classified information must not be entered. The umbrella agent will refuse to proceed if it detects classified content.

---

Install

This plugin is not yet on a public marketplace. To install during development:

git clone https://github.com/hectorir/joint-operations-planner.git ~/code/joint-planner
# Then, in Claude Code:
/plugin install ~/code/joint-planner
/reload-plugins

Quick start

# In Claude Code:
/jp                          # Front door — asks if you're starting fresh or continuing

The agent will:

1. Ask whether this is a new plan or an existing one.
2. If new: ask for a kebab-case name and plan level (estimate or bplan), then scaffold ~/joint-plans/<name>/ with all the JPP step folders, a manifest.yaml, and a context.md.
3. Run git init in the plan dir.
4. Dispatch to /jp-initiation to start Step 1.

Slash commands

Command	Purpose
/jp	Top-level router. Use when you don't know where to start.
/jp-status [plan]	Where am I? Reports step status, gaps, recommended next action.
/jp-plan	Sherpa — walk me through a step.
/jp-coach	Coach — explain doctrine, quiz me.
/jp-critique [step]	Critic — red-team a draft against the rubric.
/jp-generate [artifact]	Generator — produce a planning artifact.
/jp-initiation	JPP Step 1 — Planning Initiation.
/jp-mission-analysis	JPP Step 2 — Mission Analysis (the heaviest step).
/jp-coa-development	JPP Step 3 — COA Development.
/jp-wargame	JPP Step 4 — COA Analysis & Wargaming.
/jp-coa-comparison	JPP Step 5 — COA Comparison.
/jp-coa-approval	JPP Step 6 — COA Approval.
/jp-plan-development	JPP Step 7 — Plan or Order Development.
/jp-cog [friendly|threat]	COG / critical-factors analysis. Cross-cuts steps.
/jp-risk	Risk assessment. P × C, mitigation, civilian-leader script.
/jp-assess	Operation assessment plan. MOEs / MOPs / indicators.

Plan workspace

Each plan lives in its own directory:

~/joint-plans/<plan-name>/
├── README.md
├── manifest.yaml                 # plan state — single source of truth
├── context.md                    # plan-specific situational frame
│
├── 00_initiation/
├── 01_mission_analysis/
├── cog/                          # cross-cutting
├── 02_coas/
├── 03_wargame/
├── 04_comparison/
├── 05_approval/
├── 06_plan/
├── assess/                       # cross-cutting
└── log/
    ├── decisions.md
    ├── changes.md
    └── critiques/                # one file per /jp-critique invocation

The plan dir is git-versioned automatically. Skills auto-commit at step transitions with structured messages (step:mission-analysis complete).

Personalization

• Profile (optional): ~/.config/joint-planner/profile.md — defaults that apply across all your plans (preferences, recurring strategic-guidance citations, default plan level).
• Per-plan context (always): <plan>/context.md — the situational frame for that plan (higher HQ, AOR, stakeholders, citations for this plan).

Per-plan context overrides profile defaults on conflict.

Worked example

The plugin ships with a fully worked example: examples/exercise-pacific-cyclone/. It is a fictional UNCLASSIFIED HADR scenario (cyclone strikes the fictional Republic of Cernavia; CCMD directed to plan a JTF response in support of USAID). All seven JPP steps are scaffolded as complete and the artifacts demonstrate what each step's output looks like in finished form.

Use it to learn the workflow, browse what doctrinally-faithful artifacts look like, or copy fresh to ~/joint-plans/learn-jpp/ and walk yourself through.

How it stays doctrinally faithful

1. The Practitioner Field Guide is the canonical source. It ships at references/jp5-0-field-guide.md and is the voice every skill quotes.
2. Every skill embeds the relevant field-guide section in its prompt (or directs the agent to Read it on entry).
3. Critic mode walks an explicit rubric. The rubrics in references/rubrics/ enumerate the checks; critic mode renders PASS/WARN/FAIL with a quote, doctrinal cite, and remediation per item.
4. Step transitions are gated. A per-step skill cannot mark itself complete until critic-in-dry-run returns zero FAILs.
5. Stale propagation. Re-entering an upstream step marks downstream artifacts stale automatically. JP 5-0 stop-revise-restart discipline is structurally enforced.

Architecture

Invocation surface  (/jp slash commands · subagent_type:joint-planner)
        │
Skills layer  (one skill per command; share helpers; stateless)
        │
Knowledge layer  (jp5-0-field-guide.md · templates · checklists · rubrics)
        │
Plan workspace  (~/joint-plans/<plan>/ + manifest + git history)

See docs/superpowers/specs/2026-05-07-joint-planner-design.md for the full design.

Roadmap (Phase 2)

• A local web UI that wraps the same plan-dir model. Forms per JPP step, rendered artifacts, search.
• Optional retrieval over the source JP 5-0 PDF for verbatim citation when the field guide is silent.
• Plugin-level config (profile defaults beyond the per-plan context).
• CONPLAN-level annex generation (Levels 3+).

Status

v0.1.0 — Initial release.

License

MIT — see LICENSE.

Acknowledgements

Built on top of:

• Joint Publication 5-0, Joint Planning (Joint Chiefs of Staff, 1 December 2020).
• The Claude Code plugin system.
• Inspiration from the superpowers plugin for skill structure.