agentleFS
mmNalakammNalaka/nm-skills/skills/workflow/SKILL.md

workflow

Skill1 starsChanged 6 months agoMITView on GitHub

Use to run the full development pipeline autonomously — brainstorm through commit with two approval gates only (task intake and final commit confirmation).

71 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use build-keil, can-debug and 10 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.
  • ZhongYechengSkill

    build-keil

    当需要通过 Keil MDK 命令行编译嵌入式工程,调用自带脚本解析工程文件、执行构建并定位固件产物时使用。

    Also in 3 projects that use this

  • ZhongYechengSkill

    can-debug

    当需要调试 CAN 总线通信时使用,支持通过 USB-CAN 适配器监听、发送 CAN 帧和扫描节点。

    Also in 3 projects that use this

  • iPlug2Skill

    build-cmake

    Build an iPlug2 plugin project using CMake with Ninja, Xcode, or Visual Studio generators

    Also in 3 projects that use this

  • +9 more from ZhongYecheng

The file

Show its rules and text

We split this skill into the rules it gives an agent

22 rules. 0 are common, 19 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 55 edited versions we read keep these.

  • 19

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not proceed until user replies.

    line 47
  2. Distinctive:

    After Gate 1, run each phase in sequence.

    line 53
  3. Very specific:

    Run brainstorm rules from /nm:brainstorm but condensed:

    line 57
  4. Very specific:

    Run plan rules from /nm:plan:

    line 76
  5. Distinctive:

    Run verify check after each step

    line 97
  6. Distinctive:

    If verify fails: fix before moving to next step

    line 98
  7. Distinctive:

    Do NOT ask "continue to next step?" — auto-proceed on success

    line 99
  8. Distinctive:

    If a step fails 2 attempts: surface to user with exact error before continuing.

    line 107
  9. Distinctive:

    If Gate 1 Q3 = option 1 (full TDD):

    line 111
  10. Distinctive:

    Run tests — show pass/fail output

    line 114
  11. Distinctive:

    If tests RED after implementation: fix implementation, re-run

    line 115
  12. Distinctive:

    If Gate 1 Q3 = option 2 (tests after):

    line 118
  13. Distinctive:

    Write tests after implementation

    line 119
  14. Distinctive:

    Run once — show results

    line 120
  15. Distinctive:

    If Gate 1 Q3 = option 3: skip Phase 4.

    line 123
  16. Very specific:

    Only commit after explicit y or message edit + confirmation.

    line 174
  17. Distinctive:

    Always surface to user (break autonomous flow) when:

    line 180
  18. Distinctive:

    Do not ask questions between Gate 1 and Gate 2 unless escalation rules trigger

    line 201
  19. Distinctive:

    Do not run git commit without Gate 2 confirmation

    line 202
  20. Distinctive:

    Do not skip TDD if Gate 1 Q3 = option 1

    line 203
  21. Distinctive:

    Do not fix MEDIUM code review issues autonomously (note them, don't touch)

    line 204
  22. Distinctive:

    Do not refactor code not related to the current task

    line 205

“Common” is measured: kept by at least 6 in 10 of the 55 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

Workflow

Autonomous end-to-end development pipeline. Runs the full nm skill chain with minimal interruption:

brainstorm → plan → ts-api | go-api → tdd → code-review → commit

Two gates only: start (task + stack) and commit (final confirmation). Everything between is autonomous.

Recommended model: Sonnet (orchestration) / delegate heavy phases to sub-agents


Gate 1 — Task Intake (REQUIRED STOP)

Present this intake form and wait for answers before proceeding:

WORKFLOW START
──────────────
Task: [restate task in one sentence]

Clarify:
Q1. Stack?
  1. TypeScript (Node.js / Hono / Express)
  2. Go (net/http / Chi / Gin)

Q2. Scope?
  1. New feature (green field)
  2. Change existing feature
  3. Bug fix

Q3. Tests required?
  1. Yes — full TDD (red → green → refactor)
  2. Yes — write tests after implementation
  3. No tests needed

Reply with three numbers (e.g. "1 1 1") or describe changes.

Do not proceed until user replies.


Autonomous Pipeline

After Gate 1, run each phase in sequence. Do not pause between phases unless a phase signals ambiguity or failure.

Phase 1 — Brainstorm (auto-condensed)

Run brainstorm rules from /nm:brainstorm but condensed:

  • State understood task + assumptions (one block)
  • Identify the simplest approach
  • List top 2 risks
  • Do NOT ask clarifying questions — use Gate 1 answers as context
  • Do NOT stop for "ready to plan?" — auto-proceed to Phase 2

Output format:

[BRAINSTORM]
Task: ...
Assumptions: ...
Simplest approach: ...
Risks: ...
→ Proceeding to plan...

Phase 2 — Plan

Run plan rules from /nm:plan:

  • Build numbered steps with verify checks
  • Estimate complexity
  • Do NOT stop for approval — present plan and immediately continue

Output format:

[PLAN]
1. ... → verify: ...
2. ... → verify: ...
Complexity: Low | Medium | High
→ Proceeding to implementation...

Phase 3 — Implementation

Based on Gate 1 stack choice:

  • TypeScript → apply /nm:ts-api patterns
  • Go → apply /nm:go-api patterns

Execute plan steps one by one:

  • Run verify check after each step
  • If verify fails: fix before moving to next step
  • Do NOT ask "continue to next step?" — auto-proceed on success

Output format per step:

[IMPL] Step N/total: ...
  ✓ verify passed

If a step fails 2 attempts: surface to user with exact error before continuing.

Phase 4 — TDD

If Gate 1 Q3 = option 1 (full TDD):

  • Apply /nm:tdd Phase 1 (RED) and Phase 2 (GREEN)
  • Auto-select test cases based on implemented endpoints/functions
  • Run tests — show pass/fail output
  • If tests RED after implementation: fix implementation, re-run
  • Auto-skip refactor phase (user can run /nm:tdd separately if needed)

If Gate 1 Q3 = option 2 (tests after):

  • Write tests after implementation
  • Run once — show results
  • Fix failures

If Gate 1 Q3 = option 3: skip Phase 4.

Output format:

[TDD] Writing tests for: ...
  ✓ N tests passing

Phase 5 — Code Review

Apply /nm:code-review rules autonomously:

  • Review all files changed in this workflow
  • Fix any CRITICAL or HIGH issues found
  • List MEDIUM issues but do not fix (surface to user in Gate 2)
  • Do NOT stop for approval — auto-fix criticals and proceed

Output format:

[REVIEW]
  ✓ Fixed: [critical/high issues]
  ⚠ Noted: [medium issues for user awareness]
→ Proceeding to commit...

Gate 2 — Commit Confirmation (REQUIRED STOP)

Present full summary and wait before committing:

WORKFLOW COMPLETE — REVIEW BEFORE COMMIT
─────────────────────────────────────────
Files changed:
  [list all modified/created files]

Test results:
  [pass count / skip info]

Issues fixed:
  [list from code review]

Issues noted (not fixed):
  [list medium issues]

Commit message:
  [proposed conventional commit message]

Proceed with commit? [y / n / edit message]

Only commit after explicit y or message edit + confirmation.


Error Escalation Rules

Always surface to user (break autonomous flow) when:

  • A verify step fails twice
  • Tests fail after 2 fix attempts
  • Ambiguity requires a design decision (e.g., conflicting patterns found in codebase)
  • File conflicts or merge issues detected
  • Security concern flagged during review

Format escalation as:

[BLOCKED] Phase N — <reason>
Options:
  1. [suggested fix]
  2. [alternative]
  3. Skip this step and continue
Reply to unblock.

What NOT to do

  • Do not ask questions between Gate 1 and Gate 2 unless escalation rules trigger
  • Do not run git commit without Gate 2 confirmation
  • Do not skip TDD if Gate 1 Q3 = option 1
  • Do not fix MEDIUM code review issues autonomously (note them, don't touch)
  • Do not refactor code not related to the current task

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.