agentleFS
Dqz00116Dqz00116/skill-lib/devflow/SKILL.md

devflow

Skill22 starsChanged 7 months agoMITView on GitHub

Use when working in a DevFlow project with .devflow/ directory and gate-based step-by-step workflows

  • Installs packages

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 4 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

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

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

    If devflow is not found, install first:

    line 30
  2. Distinctive:

    When gates fail, the workflow may route based on failure count:

    line 80

“Common” is measured: kept by at least 6 in 10 of the 4 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

DevFlow

CLI-driven progressive disclosure workflow. Advance one verified step at a time.

Overview

DevFlow enforces a disciplined development process where each step must be completed and verified before advancing. It uses TOML workflow definitions, gate-based verification, and a devflow done checkpoint system.

When to Use

  • Project has .devflow/workflows/*.toml files
  • You see devflow commands in project docs
  • Task requires writing requirements, design docs, or code in stages
  • You need gate-based verification (tests must pass before advancing)

When NOT to use: Project has no .devflow/ directory — this is a standard project, use normal development.

Quick Start

devflow --help          # Verify CLI is installed

If devflow is not found, install first:

pip install agent-devflow     # From PyPI

Then start the workflow:

devflow list-workflows              # See available workflows
devflow select-workflow MODE-A      # Pick one
devflow current                     # Read current step

Core Loop (ALWAYS follow this)

1. devflow current     -> Read what to do
2. Execute the step    -> Write code, create docs, run tests
3. devflow done        -> Check gates -> advance or retry
4. Repeat until "Workflow complete!"

Core rule: Only advance through devflow done. Never skip steps. Never create files for future steps.

Commands

CommandPurpose
devflow currentShow current step instruction
devflow doneCheck gates and advance
devflow backGo back one step
devflow approve ITEMMark item as user-approved
devflow set KEY VALUESet state variable
devflow list-workflowsList available workflows
devflow select-workflow IDSelect and start a workflow
devflow runStart autonomous Ralph Loop

Gate Types

Gates block advancement until satisfied:

  • file_exists:path — File must exist
  • file_contains:path:content — File must contain text
  • command_success:{test_command} — Command exits 0
  • user_approved:ITEM — Requires devflow approve ITEM
  • state_set:var — State variable must be set

Fail Routes

When gates fail, the workflow may route based on failure count:

fails 1-2 -> retry or fallback step
fails 3+  -> escalate to human

Fail count persists across visits (resets only when gates pass).

Built-in Workflows

WorkflowSteps
MODE-A (Feature)req-create -> req-approve -> brainstorm -> write-plan -> implement-sdd -> code-review -> test-run -> verify -> finish
MODE-B (Debug)debug-root-cause -> debug-pattern -> debug-hypothesis -> debug-fix -> debug-question -> debug-verify -> debug-finish -> (auto -> MODE-A:write-plan)

Cross-workflow references use WORKFLOW-ID:STEP-ID format.

5 Iron Laws

#LawSeverity
1Read using-superpowers skill firstRigid
2TDD: Test before codeRigid
3Verify before claimRigid
4Root cause before fixStrong
5Debug -> Repeat full cycleStrong

Common Mistakes

MistakeFix
Writing code before devflow currentStop. Run devflow current first
Creating files for future stepsOnly create what the current step asks for
Skipping devflow doneAll advancement must go through devflow done
Ignoring gate failuresFix the issue, then devflow done again

Project Context

Language: {{ project.language }} Stack: {{ project.stack }} Version: {{ project.version }}

Constraints

{% if constraints.zero_warnings %}- ZERO warnings — All builds must have zero warnings{% endif %} {% if constraints.zero_mocks %}- ZERO mocks — Use real instances only in tests{% endif %} {% if constraints.nullable %}- Nullable enabled — Use nullable reference types{% endif %}

Project Structure

{{ project.name }}/
├── .devflow/              # DevFlow configuration
│   ├── workflows/         # Workflow definitions (TOML)
│   ├── prompts/           # Reusable prompt files
│   ├── config.toml        # Project config
│   └── state.toml         # Current state (gitignored)
├── {{ paths.src }}/       # Source code
├── {{ paths.tests }}/     # Tests
├── {{ paths.docs }}/      # Documentation
│   ├── requirements/      # REQ files
│   ├── features/          # FEAT files
│   ├── superpowers/specs/ # Design docs
│   ├── superpowers/plans/ # Implementation plans
│   ├── evidence/          # Verification evidence
│   ├── debug/             # Debug analysis
│   └── completion/        # Workflow completion summaries
└── SKILL.md               # This file

Commands Reference

# Build
{{ commands.build }}

# Test
{{ commands.test }}
{{ commands.test_unit }}
{{ commands.test_integration }}

# Lint
{{ commands.lint }}

Generated by DevFlow v2.0

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.