planning-workflow
Skill3 starsChanged 3 months agoMITView on GitHub
Plan implementation tasks by validating the problem, reconciling DESIGN.md, searching past solutions, calibrating research depth, analyzing edge cases via SpecFlow, and generating living plans with checkable criteria. Use when starting implementation planning for a task or issue.
6 other projects use a version of this.
Adopted by
Most recently taken up by
- DITlieD/ELAI-archivetheir own version, first seen Sep 5, 2026
- 2702207741-dev/agent-skills-pipelinetheir own version, first seen Jul 9, 2026
- Dicklesworthstone/agent_flywheel_clawdbot_skills_and_integrationstheir own version, first seen Jun 19, 2026
- indubitably-ai/indubitably-skillstheir own version, first seen Apr 17, 2026
- HyunKN/harness-kittheir own version, first seen Apr 7, 2026
- LazyWorkshopCreate/agent-based-workflowtheir own version, first seen Mar 25, 2026
Other versions6 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 6
- Dicklesworthstone/agent_flywheel_clawdbot_skills_and_integrations★ 73same name
- DITlieD/ELAI-archive★ 24same name
- LazyWorkshopCreate/agent-based-workflow★ 9same name
- 2702207741-dev/agent-skills-pipeline★ 1same name
- indubitably-ai/indubitably-skills★ 1same name
- HyunKN/harness-kit★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository41 other files it gives its agents.
- cadence-goalsSkill
- ci-feedback-loopSkill
- compoundSkill
- decision-rights-doctrineSkill
- diataxisSkill
- dispatch-gateSkill
- drift-checkSkill
- family-checkSkill
- fix-then-re-review-ladderSkill
- github-projectsSkill
- gitSkill
- goal-treeSkill
- handoffSkill
- improvement-loopSkill
- kb-captureSkill
- kb-compileSkill
- kb-coreSkill
- kb-lessons-bridgeSkill
- kb-lintSkill
- l1-supervisorSkill
- l2-supervisorSkill
- lessons-learnedSkill
- lN-lifecycle-doctrineSkill
- lN-review-doctrineSkill
- medical-recordsSkill
- operator-interview-doctrineSkill
- operator-interviewSkill
- operator-reviewSkill
- pr-review-fanoutSkill
- pull-taskSkill
- ralph-wiggumSkill
- reviewSkill
- root-cause-depthSkill
- scan-documentSkill
- self-improvementSkill
- self-verifySkill
- skillifySkill
- task-workflowSkill
- testing-without-mocksSkill
- vault-cloudSkill
- verify-inherited-premiseSkill
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 9 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 5 edited versions we read keep these.
9
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 53
When: Always — first step of any planning session, before solution search.
- Distinctive: line 63
When: After problem validation — reconciles workspace DESIGN.md before solution search.
- Distinctive: line 71
When: After DESIGN.md reconciliation — evaluates whether to skip phases 3-6.
- Distinctive: line 79
When: After fast path gate (full path only) — searches for relevant prior solutions.
- Distinctive: line 87
When: After solution search — decides whether to invoke external research.
- Distinctive: line 95
When: After research, before finalizing plan.
- Distinctive: line 103
When: After SpecFlow analysis — determines plan structure.
- Distinctive: line 119
When: After plan generation — propagates superseding design decisions to downstream artifacts.
- Distinctive: line 127
When invoked, run all phases sequentially.
- Very specific: line 136
When Plan Mode is available, EnterPlanMode wraps the pipeline (phases 1–7 draft ephemerally), ExitPlanMode presents the assembled draft for …
- Very specific: line 261
Graceful degradation: Every phase handles missing inputs without failing.
“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Planning Workflow
Structured planning that validates the problem and mines past knowledge before investing in new research. Produces living plans with checkable acceptance criteria.
Plan Mode Front-End
Plan Mode (EnterPlanMode → interactive drafting → ExitPlanMode) is the
interactive drafting front-end that brackets the phase pipeline below. The draft
is assembled ephemerally inside Plan Mode (no durable writes); only on user
approval does the materialization step write the durable DESIGN.md/PLAN.md
and seed native TaskCreate/TaskList. Plan Mode is ephemeral (it resets on
acceptEdits) — the durable docs and seeded native tasks remain authoritative.
See operations/plan-mode-frontend.md. If Plan Mode is unavailable, run the
pipeline directly and write durable artifacts as before.
Overview
The planning workflow runs these phases in order:
- Problem validation — validate who, what pain, current workflow, and success criteria; for document tasks, gate on content correctness before proceeding
- DESIGN.md reconciliation — compare workspace DESIGN.md against problem validation, repo docs, and strategic docs
- Solution search — QMD hybrid search over the Obsidian vault for relevant prior notes (DD4)
- Research gating — decide if external research is needed based on risk
- SpecFlow analysis — walk user flows to find edge cases and audit the layer each requirement belongs to
- Detail level selection — choose plan depth based on complexity
- Plan generation — produce markdown plan with checkable criteria
- ADR propagation — propagate superseding decisions to strategic docs, issues, and project board
Operations
1. Problem Validation
When: Always — first step of any planning session, before solution search.
Implementation: Load operations/problem-validation.md
Quick summary: Scans the task description for four dimensions (user, pain, current workflow, success criteria). The elicitation logic (question taxonomy/funnel, ask-vs-default rule, stop conditions, AskUserQuestion rules) is sourced from the operator-interview-doctrine skill — this phase is a thin caller that binds the doctrine to the four dimensions. Extracts what's covered, infers what's implied, and interviews the user for anything missing. Then classifies the task as document/plan vs. code implementation using keyword heuristics (with confidence fallback to user). For document tasks, presents an interactive content gate — "Content is correct" proceeds; "Needs revision" stops the workflow. Produces a "Problem Validation" section that grounds all downstream phases.
Explicit entry point: For a standalone operator interview outside the planning pipeline, use the /operator-interview <topic> command, which runs the same operator-interview-doctrine.
2. DESIGN.md Reconciliation
When: After problem validation — reconciles workspace DESIGN.md before solution search.
Implementation: Load operations/designmd-reconciliation.md
Quick summary: Three-layer validation of workspace DESIGN.md against problem validation findings, repo-level docs (CLAUDE.md, docs/), and strategic docs (strategic meta-repo REQUIREMENTS.md, ARCHITECTURE.md, PROJECT.md). Surfaces contradictions and gaps for user approval. Bidirectional conflict resolution for strategic docs — changes may flow down (update DESIGN.md) or up (flag strategic doc update). Produces a "DESIGN.md Reconciliation" section for the plan.
2a. Fast Path Gate
When: After DESIGN.md reconciliation — evaluates whether to skip phases 3-6.
Implementation: Load operations/fast-path-gate.md
Quick summary: Checks three deterministic criteria: (1) task has explicit, testable acceptance criteria, (2) task targets known, locatable code, (3) DESIGN.md reconciliation is clean (no unresolved contradictions). All three must pass for fast path. On fast path, skips directly to plan generation with Minimal detail level. On full path, continues to phase 3.
3. Solution Search
When: After fast path gate (full path only) — searches for relevant prior solutions.
Implementation: Load operations/solution-search.md
Quick summary: Delegates a QMD query (hybrid BM25 + vector + reranker + HyDE) against the configured vault collection to an out-of-context Explore subagent (Task tool, subagent_type: Explore), which follows task-workflow/references/solution-search.md for the canonical invocation and fail-open protocol and returns the compiled "Prior Solutions" section. The planning session consumes that section as context for downstream phases; it does not run qmd directly. Per DD4, no longer reads docs/solutions/.
4. Research Gating
When: After solution search — decides whether to invoke external research.
Implementation: Load operations/research-gating.md
Quick summary: Classifies task risk (high/low/uncertain), evaluates local knowledge strength, and decides whether external research is needed. High-risk topics always get external research; low-risk with strong local knowledge skips it.
5. SpecFlow Analysis
When: After research, before finalizing plan.
Implementation: Load operations/specflow-analysis.md
Quick summary: Systematically walks all user/operational flows, maps decision points, enumerates error states and edge cases, identifies specification gaps, audits where each requirement should run to catch layer mismatches before implementation, and generates acceptance criteria from findings.
6. Detail Level Selection
When: After SpecFlow analysis — determines plan structure.
Implementation: Load operations/detail-level.md
Quick summary: Scores complexity signals (flows, edge cases, spec gaps, files affected) and selects Minimal, More, or A Lot detail level. Determines which sections the plan generator includes.
7. Plan Generation
When: Final phase — assembles the plan document.
Implementation: Load operations/plan-generation.md
Quick summary: Generates PLAN.md using the selected detail level template, populated with findings from all prior phases. Acceptance criteria use [ ]/[x] checkboxes for progress tracking during implementation. Standard documentation and demo criteria are always included (with deduplication against task-specific criteria).
8. ADR Propagation
When: After plan generation — propagates superseding design decisions to downstream artifacts.
Implementation: Load operations/adr-propagation.md
Quick summary: Consumes upstream flags from DESIGN.md reconciliation. For strategic docs (PROJECT.md, REQUIREMENTS.md, ARCHITECTURE.md), proposes specific edits and applies with user approval. For GitHub issues and project board items, generates actionable checklist entries. Skips cleanly when no upstream flags exist.
End-to-End Flow
When invoked, run all phases sequentially. Each phase produces a section that feeds into the next. After phase 2, a fast path gate may skip phases 3-6 for qualifying tasks (see "Fast Path" below).
Input
- Task description (from issue, DESIGN.md, or user input)
- Repository context (working directory, available files)
Phase Pipeline
When Plan Mode is available, EnterPlanMode wraps the pipeline (phases 1–7 draft
ephemerally), ExitPlanMode presents the assembled draft for approval, and the
materialization step writes durable DESIGN.md/PLAN.md + seeds native Tasks
(see operations/plan-mode-frontend.md).
EnterPlanMode (read-only drafting; no durable writes)
│
▼
Task Description
│
▼
┌─────────────┐
│ 1. Problem │──→ Problem Validation section
│ Validation │ (user, pain, current workflow, success criteria)
│ │ + task type classification
└─────────────┘
│
├── if document task → Interactive Content Gate
│ ├── "Content is correct" → continue
│ └── "Needs revision" → STOP (revise and re-run)
│
▼
┌─────────────┐
│ 2. DESIGN.md │──→ DESIGN.md Reconciliation section
│ Reconcile │ (3-layer check: validation, repo docs, strategic docs)
└─────────────┘
│
▼
┌─────────────┐
│ 2a. Fast │──→ Fast Path Gate section
│ Path Gate │ (criteria check: acceptance criteria + known code + low risk)
└─────────────┘
│
├── all criteria pass → FAST PATH (skip to phase 7)
│
▼ (any criterion fails → full path)
┌─────────────┐
│ 3. Solution │──→ Prior Solutions section
│ Search │ (matched solutions + critical patterns)
└─────────────┘
│
▼
┌─────────────┐
│ 4. Research │──→ Research Decision section
│ Gating │ (risk category + local knowledge → skip/research)
└─────────────┘
│
├── if "research externally" → conduct web search / doc lookup
│
▼
┌─────────────┐
│ 5. SpecFlow │──→ SpecFlow Analysis section
│ Analysis │ (flows, edge cases, gaps, generated criteria)
└─────────────┘
│
▼
┌─────────────┐
│ 6. Detail │──→ Plan Detail Level section
│ Level │ (Minimal / More / A Lot)
└─────────────┘
│
▼
┌─────────────┐
│ 7. Plan │──→ PLAN.md
│ Generation │ (living plan with checkable criteria)
└─────────────┘
│
▼
┌─────────────┐
│ 8. ADR │──→ ADR Propagation section
│ Propagation │ (strategic doc edits + issue/board checklist)
└─────────────┘
│
▼
ExitPlanMode ──→ present assembled draft for approval
│
▼ (approved)
MATERIALIZE ──→ write durable DESIGN.md/PLAN.md (phase 7 step 5)
+ seed native TaskCreate/TaskList from criteriaOutput
PLAN.md written to the workspace root, containing:
- Plan body (sections per detail level)
- Acceptance criteria with
[ ]checkboxes - Planning context appendix (outputs from phases 1-7)
- ADR propagation section (strategic doc changes applied + issue/board checklist)
Fast Path
An alternative flow for tasks that don't benefit from phases 3-6. The fast path produces the same PLAN.md output format as the full path, using Minimal detail level.
When It Applies
The fast path gate runs after phase 2 (DESIGN.md reconciliation) and checks three deterministic criteria. All three must pass:
- Explicit acceptance criteria — the task description contains at least one testable, pass/fail statement
- Known code target — the task targets files or modules that exist in the repo and can be located (via Glob/Grep)
- Low risk — DESIGN.md reconciliation produced no unresolved contradictions
What Gets Skipped
| Phase | Full path | Fast path |
| 1. Problem validation | Runs | Runs |
| 2. DESIGN.md reconciliation | Runs | Runs |
| 2a. Fast path gate | Runs | Runs |
| 3. Solution search | Runs | Skipped |
| 4. Research gating | Runs | Skipped |
| 5. SpecFlow analysis | Runs | Skipped |
| 6. Detail level selection | Runs | Skipped (defaults to Minimal) |
| 7. Plan generation | Runs | Runs |
Implementation
Load operations/fast-path-gate.md for criteria definitions and execution details.
Full path remains the default. The fast path gate evaluates criteria and routes automatically — no user decision required.
Common Patterns
Query term extraction: Derive search terms from the issue title, description, error messages, and component names. Use 3-5 specific terms rather than broad keywords.
Graceful degradation: Every phase handles missing inputs without failing. If QMD is not installed or $QMD_COLLECTION is unset, solution-search logs and continues — planning never blocks on retrieval (see task-workflow/references/solution-search.md).
Progressive Disclosure
Load only what you need:
operations/plan-mode-frontend.md— Plan Mode front-end:EnterPlanMode→ ephemeral drafting →ExitPlanMode→ materialize durable docs + seed native Tasksoperations/problem-validation.md— Problem validation, user interview, task type classification, and interactive content gateoperations/designmd-reconciliation.md— DESIGN.md three-layer reconciliationoperations/fast-path-gate.md— Fast path criteria evaluation and routingoperations/solution-search.md— Solution search implementation detailsoperations/research-gating.md— Research gating decision logicoperations/specflow-analysis.md— SpecFlow edge case analysisoperations/detail-level.md— Detail level selection logicoperations/plan-generation.md— Plan generation with checkable criteriaoperations/adr-propagation.md— Post-ADR propagation to strategic docs, issues, and board
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

