agentleFS
ekirtonekirton/Poule/.claude/skills/sdd/SKILL.md

sdd

Skill0 starsChanged 8 months agoApache-2.0View on GitHub

Run the Spec-Driven Development pipeline for a feature or change. Walks through SDD stages in order with enforced layer boundaries and feedback loops. Use when asked to build a feature end-to-end, propagate a change, fix a bug through SDD, or follow the SDD process. Accepts an optional starting layer as the first argument (e.g., "/sdd specification fix the caching bug").

22 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use bdd, migrate and 3 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • TashanGKDSkill

    bdd

    行为驱动开发(BDD)流程助手。用 Given-When-Then 描述验收标准,包含 Feature 文件管理、步骤定义实现。用户提到 BDD、验收测试、Gherkin 时使用。

    Also in 3 projects that use this

  • tuistSkill

    migrate

    Migrates existing Xcode projects to Tuist generated workspaces with build and run validation, external dependency mapping, and migration checklists. Use when adopting Tuist for an existing app or converting a hand-edited Xcode project to generated projects.

    Also in 4 projects that use this

  • BlockSkill

    plan

    Draft a TDD, PRD, or lightweight plan in the conversation without committing it.

    Also in 3 projects that use this

  • Hugging FaceSkill

    implement

    Make tests pass. Invoke after /write-tests produces failing tests.

    Also in 3 projects that use this

  • GoogleSkill

    tdd

    Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.

    Also in 4 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

22 rules. 0 are common, 12 are distinctive, and 10 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 12

    Distinctive

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If propagating from a user request that clearly defines the requirements, write or update the PRD in doc/requirements/ autonomously.

    line 43
  2. Distinctive:

    If the requirements already exist and are sufficient, confirm with the user and skip to Stage 2.

    line 44
  3. Distinctive:

    If the change requires judgment (e.g., scope decisions, priority trade-offs), present options with analysis and a recommendation.

    line 45
  4. Very specific:

    If propagating from Stage 1, propagate to doc/features/ autonomously.

    line 50
  5. Distinctive:

    If this is the starting stage, or if the change requires judgment beyond what the requirements prescribe, present options with analysis and …

    line 51
  6. Distinctive:

    If a problem with the requirements is detected, do not edit requirements — surface the issue to the user and stop.

    line 52
  7. Very specific:

    Read doc/architecture/data-models/expression-tree.md and doc/architecture/data-models/index-entities.md.

    line 57
  8. Distinctive:

    If this is the starting stage, or if the change involves design decisions not prescribed by the feature doc, present options with analysis …

    line 59
  9. Distinctive:

    If a problem is found in upstream documents, do not edit them — surface the issue to the user and stop.

    line 60
  10. Distinctive:

    Read the parent architecture document.

    line 65
  11. Very specific:

    If a problem is identified with the architecture, write a detailed description to doc/architecture/feedback/ and stop.

    line 67
  12. Very specific:

    Read the blast radius from $CLAUDE_PROJECT_DIR/sdd-blast-radius if it exists.

    line 75
  13. Distinctive:

    Write or update tests for the new/changed specifications within the blast radius.

    line 77
  14. Distinctive:

    Do not change specifications.

    line 78
  15. Very specific:

    Run python -m pytest test/ -x -q.

    line 79
  16. Distinctive:

    Write the implementation to make tests pass.

    line 81
  17. Very specific:

    Run python -m pytest test/ -x -q after each significant change.

    line 82
  18. Very specific:

    Write to doc/architecture/feedback/, notify the user, and stop.

    line 92
  19. Very specific:

    Remove $CLAUDE_PROJECT_DIR/sdd-blast-radius if it exists.

    line 98
  20. Very specific:

    Check off completed tasks in tasks/ (update - [ ] to - [x]).

    line 99
  21. Distinctive:

    If any feedback files still exist, notify the user and stop.

    line 100
  22. Distinctive:

    Do not make a PR — the user decides when the branch is ready.

    line 102

“Common” is measured: kept by at least 6 in 10 of the 13 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Spec-Driven Development Pipeline

Execute the SDD pipeline for: $ARGUMENTS

Determine the starting stage

Parse the first word of the arguments. If it matches a layer name, start at that stage and skip earlier stages. Otherwise, start at Stage 1.

First wordStart at
requirementsStage 1
featuresStage 2
architectureStage 3
specificationStage 4
testsStage 5
implementationStage 5
(anything else)Stage 1 (full pipeline)

Note: implementation maps to Stage 5 (tests first), enforcing TDD. To skip tests entirely, use the /implementation phase command directly instead of /sdd.

Work through the stages below starting from the determined stage. At each stage, set the phase to enforce layer boundaries. Do not invent requirements or infer unnecessary details — ask the user when ambiguities exist.

Autonomy rules

Human-in-the-loop layers (doc/requirements/, doc/features/, doc/architecture/):

  • Propagating downward from the stage above (e.g., requirements→features, features→architecture): proceed autonomously.
  • Originating changes (writing initial content, fixing a gap found during triage, or making changes NOT derived from the stage immediately above): present 2-3 options with critical analysis and a recommendation. Wait for the user to approve before editing.

Autonomous layers (specification/, test/, src/):

  • Proceed without human intervention.
  • After 3 feedback cycles, stop and present the situation to the user. This prevents infinite loops. Summarize what was attempted, what keeps failing, and ask for direction.

Stage 1: Requirements

  1. Run: echo "requirements" > $CLAUDE_PROJECT_DIR/sdd-layer
  2. If propagating from a user request that clearly defines the requirements, write or update the PRD in doc/requirements/ autonomously.
  3. If the requirements already exist and are sufficient, confirm with the user and skip to Stage 2.
  4. If the change requires judgment (e.g., scope decisions, priority trade-offs), present options with analysis and a recommendation. Wait for approval.

Stage 2: Features

  1. Run: echo "features" > $CLAUDE_PROJECT_DIR/sdd-layer
  2. If propagating from Stage 1, propagate to doc/features/ autonomously.
  3. If this is the starting stage, or if the change requires judgment beyond what the requirements prescribe, present options with analysis and a recommendation. Wait for approval.
  4. If a problem with the requirements is detected, do not edit requirements — surface the issue to the user and stop.

Stage 3: Architecture

  1. Run: echo "architecture" > $CLAUDE_PROJECT_DIR/sdd-layer
  2. Read doc/architecture/data-models/expression-tree.md and doc/architecture/data-models/index-entities.md.
  3. If propagating from Stage 2, propagate to doc/architecture/ autonomously.
  4. If this is the starting stage, or if the change involves design decisions not prescribed by the feature doc, present options with analysis and a recommendation. Wait for approval.
  5. If a problem is found in upstream documents, do not edit them — surface the issue to the user and stop.

Stage 4: Specifications

  1. Run: echo "specification" > $CLAUDE_PROJECT_DIR/sdd-layer
  2. Read the parent architecture document.
  3. Propagate architecture down to specification/.
  4. If a problem is identified with the architecture, write a detailed description to doc/architecture/feedback/ and stop. Notify the user — architecture changes require human approval.
  5. Record the blast radius — write changed spec filenames to $CLAUDE_PROJECT_DIR/sdd-blast-radius: echo "specification/channels.md specification/storage.md" > $CLAUDE_PROJECT_DIR/sdd-blast-radius

Stage 5: Tests and Implementation

This stage follows TDD: write failing tests first, then implement until tests pass. Track feedback cycles (any return to step 1 or 3 from a feedback resolution counts as one cycle).

  1. Read the blast radius from $CLAUDE_PROJECT_DIR/sdd-blast-radius if it exists.
  2. Run: echo "tests" > $CLAUDE_PROJECT_DIR/sdd-layer
  3. Write or update tests for the new/changed specifications within the blast radius.
  4. Do not change specifications. If a spec problem is found, go to step 9.
  5. Run python -m pytest test/ -x -q. New tests covering the change should fail. Existing tests may pass or fail.
  6. Run: echo "implementation" > $CLAUDE_PROJECT_DIR/sdd-layer
  7. Write the implementation to make tests pass. Do not change tests or specifications.
  8. Run python -m pytest test/ -x -q after each significant change. When all tests pass, proceed to Completion.
  9. If a spec problem is found (during tests or implementation):
    • Run: echo "specification" > $CLAUDE_PROJECT_DIR/sdd-layer
    • Fix the specification. Delete the feedback file.
    • Increment feedback cycle count. Return to step 2.
  10. If a test problem is found (during implementation):
    • Run: echo "tests" > $CLAUDE_PROJECT_DIR/sdd-layer
    • Fix the test. Delete the feedback file.
    • Increment feedback cycle count. Return to step 6.
  11. If an architecture problem is found:
    • Write to doc/architecture/feedback/, notify the user, and stop.
  12. After 3 feedback cycles, stop and present the situation to the user: summarize what was attempted, what keeps failing, and ask for direction.

Completion

  1. Run: echo "free" > $CLAUDE_PROJECT_DIR/sdd-layer
  2. Remove $CLAUDE_PROJECT_DIR/sdd-blast-radius if it exists.
  3. Check off completed tasks in tasks/ (update - [ ] to - [x]).
  4. If any feedback files still exist, notify the user and stop.
  5. Report what was done and which files were changed.
  6. Do not make a PR — the user decides when the branch is ready.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.