By Tsadoq
Deep planning for non-trivial work: triangulated research, AskUserQuestion-driven decisions, deep web research per branch, and an AI-consumable plan file in your repo.
Deep-plan codebase explorer. Parallel breadth-first search for existing patterns, target files, and similar features in the user's project. Read-only. Used in Phase 1 of /deep-plan.
Deep-plan adversarial critic. Tries to refute a synthesized plan before the user approves it: hunts for missing tasks, wrong dependencies, code tasks with no tests, decisions contradicted by research, and untested assumptions. Read-only. Used in Phase 4.6 of /deep-plan.
Drafts a perspective-flavoured implementation plan section (simplicity, performance, maintainability, minimal-diff, security). Used in Phase 4 fan-out of /deep-plan. Read-only.
Deep-plan focused research for a single chosen option. Reads official docs, validates the approach, surfaces gotchas and version constraints, returns a citation-rich dossier. Read-only. Used in Phase 3 of /deep-plan.
Deep-plan light web reconnaissance. Fast WebSearch sweep for library names, common patterns, and version landscape. No deep doc reading. Read-only. Used in Phase 1 of /deep-plan.
Executes a finalized /deep-plan plan file. Parses the plan's ## Tasks into harness tasks (one TaskCreate each), wires Depends on into addBlockedBy, then drives a test-first implementation loop task by task in dependency order. Invoke after a /deep-plan plan is approved and you are ready to build it, e.g. "implement the plan" or "/deep-plan:deep-plan-execute <plan-file>".
Triangulated, decision-surfacing planning for non-trivial tasks. Fans research three ways in parallel, surfaces every meaningful sub-decision as an AskUserQuestion with 3 to 5 options, runs targeted deep web research per chosen option, and produces an AI-consumable plan file with TDD-embedded tasks. Triggers on the slash command /deep-plan only.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A personal Claude Code skill for deep, co-authored planning of non-trivial work. It fans research three ways in parallel, surfaces every meaningful sub-decision as a multi-option AskUserQuestion, runs targeted deep web research per chosen option, runs an adversarial critique pass that tries to refute the plan before you approve it, and produces an AI-consumable plan file inside your repo. A companion /deep-plan:deep-plan-execute command then turns that plan into real harness tasks and drives a test-first implementation loop.
The user is a co-author of the plan, not a reviewer. The skill never silently picks between meaningful options. A depth: argument scales how hard it works, from a quick single pass to an exhaustive multi-wave run.
flowchart TD
Start(["/deep-plan [optional slug hint]"]) --> P0
P0[Phase 0: Bootstrap<br/>git toplevel, plans_dir, session state,<br/>stale-draft detection]
P0 --> P1[Phase 1: Parallel Triangulation]
P1 --> A1[dp-explore-codebase<br/>haiku, parallel]
P1 --> A2[dp-research-shallow<br/>haiku, parallel]
P1 --> A3[dp-source-ingest<br/>sonnet, conditional]
A1 --> CP1
A2 --> CP1
A3 --> CP1
CP1{Checkpoint 1<br/>scope confirm via AskUserQuestion}
CP1 -->|reframe| P1
CP1 -->|confirm| P2
P2[Phase 2: Decision Surfacing<br/>draft plan born, decisions appended live,<br/>sequential AskUserQuestion in dependency order]
P2 --> P3[Phase 3: Targeted Deep Research<br/>parallel, capped at 4]
P3 --> A4[dp-research-deep<br/>sonnet, one per decision branch]
A4 --> P3a{contradiction?}
P3a -->|yes, re-ask| P2
P3a -->|no| P4
P4[Phase 4: Synthesis & Verification<br/>perspective fan-out + sandbox POCs]
P4 --> A5[dp-plan-perspective<br/>inherit, parallel up to 3]
A5 --> P46[Phase 4.6: Adversarial critique]
P46 --> A6[dp-plan-critic<br/>inherit, refutes the plan]
A6 --> P46d{material gaps?}
P46d -->|fix inline| P4
P46d -->|reverses a decision| P2
P46d -->|clean| CP2
CP2{Checkpoint 2<br/>walk plan via AskUserQuestion,<br/>THE approval gate}
CP2 -->|refine task| P4
CP2 -->|change decision| P2
CP2 -->|approve| P5[Phase 5: in-place archive split<br/>+ post-approval handoff]
P5 --> Compact["Recommend /compact<br/>user triggers manually"]
Compact --> Exec["/deep-plan:deep-plan-execute<br/>load_tasks.py -> TaskCreate -> addBlockedBy -> TDD loop"]
In Claude Code:
/plugin marketplace add tsadoq/claude-better-plan
/plugin install deep-plan@claude-better-plan
Then in any project:
/deep-plan add a rate limiter to the API
Optional arguments (parsed from the prompt; order-free):
/deep-plan slug:rate-limiter depth:exhaustive add a rate limiter to the API
depth: shallow | standard | exhaustive -- scales fan-out and effort. shallow runs explore + shallow research only, skips deep research, one perspective, one quick critique pass. standard (default) is the full workflow with one critique loop. exhaustive runs multiple research waves, three perspectives, and loops the critique until it finds nothing material (cap 3 rounds).slug: my-name -- an explicit archive-slug hint; otherwise the slug is derived from the topic.After approval and /compact, hand the plan to implementation:
/deep-plan:deep-plan-execute # newest plan in the project plans_dir
/deep-plan:deep-plan-execute path/to/plan.md
It parses the plan's ## Tasks, creates one harness task per task (TaskCreate), wires Depends on into addBlockedBy (TaskUpdate), then implements each task test-first in dependency order. It refuses to start while ## Open questions is non-empty. Requires Claude Code >= v2.1.142 for the Task dependency API.
To install from a local checkout while developing:
/plugin marketplace add /absolute/path/to/claude-better-plan
/plugin install deep-plan@claude-better-plan
This repo is a Claude Code marketplace that ships exactly one plugin (deep-plan). The repo root is also the plugin root. Runtime data lives at $XDG_STATE_HOME/deep-plan/ (default ~/.local/state/deep-plan/) and is never git-tracked.
npx claudepluginhub tsadoq/claude-better-plan --plugin deep-planComplete creative writing suite with 10 specialized agents covering the full writing process: research gathering, character development, story architecture, world-building, dialogue coaching, editing/review, outlining, content strategy, believability auditing, and prose style/voice analysis. Includes genre-specific guides, templates, and quality checklists.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Comprehensive startup business analysis with market sizing (TAM/SAM/SOM), financial modeling, team planning, and strategic research
v9.44.1 — Patch release for Gemini environment/version detection and qwen auth gating. Run /octo:setup.
Comprehensive .NET development skills for modern C#, ASP.NET, MAUI, Blazor, Aspire, EF Core, Native AOT, testing, security, performance optimization, CI/CD, and cloud-native applications
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses