Skip to content
Back to search

DevMethodology

SkillWorks with Claude CodeGitHub

Spec-first development methodology for AI coding agents: Claude Code skills, QC agent, cross-vendor review and a VS Code setup

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/omnizenraj--devmethodology/badge.svg)](/listings/omnizenraj--devmethodology)

HTML

<a href="/listings/omnizenraj--devmethodology"><img src="/api/listings/omnizenraj--devmethodology/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

git clone <this repo> ~/DEV/DevMethodology
cd ~/DEV/DevMethodology
./install.sh --dry-run --all   # see what would change
./install.sh                   # skills, agent, command and scripts
./install.sh --vscode          # merge VS Code settings and keybindings
./install.sh --extra ~/work/work-skills   # also link a second skills folder

The installer links one copy of each skill into `~/.claude/skills/`, and into `~/.codex/skills/`, `~/.agents/skills/` and `~/.cursor/skills/` when those folders exist. Edit a skill once and every tool sees the change. Existing files are never overwritten.

In each project, copy `templates/ProjectAGENTS.md` to `AGENTS.md`, fill in its critical paths and commands, and link `CLAUDE.md` to it:

ln -s AGENTS.md CLAUDE.md

Claude Code reads `CLAUDE.md`. Codex, Cursor and most other agent tools read `AGENTS.md`.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 79 lines
  1. # Dev Methodology
  2. A spec-first method for building software with AI coding agents, packaged as Claude Code skills, a QC agent and a commit command, with a VS Code setup to match.
  3. ## The method
  4. 1. An engineer writes the brief: the objective, user stories with testable acceptance criteria, and the constraints.
  5. 2. The agent writes a design specification before any code. Every design element is stamped with the proven patterns it uses, or marked `[NOVEL]` with a reason.
  6. 3. The specification is reviewed in proportion to its risk tier:
  7. - **Trivial:** no specification.
  8. - **Standard:** a six-section specification, a quick QC pass, and a cross-vendor review of the novel elements.
  9. - **Critical:** a full specification and the Long Review, ending with a full cross-vendor review.
  10. 4. A person approves the specification.
  11. 5. The code is built to the specification and checked by the project's own tests, lint and type checks, then QC, before anything is committed.
  12. Tiers only go up automatically. Only a person can lower one.
  13. ## What is in the kit
  14. | Path | What it is |
  15. |---|---|
  16. | `skills/design-spec` | The specification lifecycle: create, update, review, approve |
  17. | `skills/ai-peer-review` | Cross-vendor review: runs a second vendor's model (GPT through the Codex CLI by default) read-only and renders a review document |
  18. | `skills/add-issue` | Creates a GitHub issue for an approved specification |
  19. | `skills/session-memory-save` | Saves session context so work survives `/clear` |
  20. | `agents/quality-control-enforcer.md` | Same-vendor QC for specifications and code |
  21. | `commands/commit.md` | `/commit`, the formal commit command |
  22. | `rules/design-spec.md` | The rules: tiers, required specification sections, pattern stamping, build steps |
  23. | `patterns/PatternsLibrary.md` | The pattern library used for stamps |
  24. | `scripts/review_tools.py` | Lints and renders the peer-review document |
  25. | `templates/ProjectAGENTS.md` | A starting `AGENTS.md` for a project that uses the method |
  26. | `templates/work-skills/` | Placeholder skills for an employer's own git flow, Jira, Kubernetes test clusters, uv and `just` recipes |
  27. | `vscode/` | Settings, keybindings and extension list |
  28. | `scripts/claude-code.command` | Terminal launcher that returns to a shell when Claude Code exits |
  29. | `scripts/patch-claude-vscode` | Keeps the Claude Code extension's terminal open after `/exit` |
  30. ## Install
  31. ```bash
  32. git clone <this repo> ~/DEV/DevMethodology
  33. cd ~/DEV/DevMethodology
  34. ./install.sh --dry-run --all # see what would change
  35. ./install.sh # skills, agent, command and scripts
  36. ./install.sh --vscode # merge VS Code settings and keybindings
  37. ./install.sh --extra ~/work/work-skills # also link a second skills folder
  38. ```
  39. The installer links one copy of each skill into `~/.claude/skills/`, and into `~/.codex/skills/`, `~/.agents/skills/` and `~/.cursor/skills/` when those folders exist. Edit a skill once and every tool sees the change. Existing files are never overwritten.
  40. In each project, copy `templates/ProjectAGENTS.md` to `AGENTS.md`, fill in its critical paths and commands, and link `CLAUDE.md` to it:
  41. ```bash
  42. ln -s AGENTS.md CLAUDE.md
  43. ```
  44. Claude Code reads `CLAUDE.md`. Codex, Cursor and most other agent tools read `AGENTS.md`.
  45. ## Work-specific skills
  46. `templates/work-skills/` holds placeholder skills for how one organisation works. Copy the folder into a repository your employer owns, fill it in from the team's docs, and install it alongside this kit with `--extra`. See its README.
  47. ## Recommended plugins
  48. These come from their publishers, so they are not copied here. Install them in Claude Code to get the latest versions:
  49. - Anthropic's `feature-dev` plugin, for its `code-explorer`, `code-architect` and `code-reviewer` agents: `/plugin install feature-dev@claude-plugins-official`
  50. - `superpowers`, for planning, test-driven development, systematic debugging and verification before completion: `/plugin install superpowers@claude-plugins-official`
  51. ## Models
  52. The skills work with any capable model. The QC agent asks for Opus. The cross-vendor review needs a second vendor's model, by default GPT through the [Codex CLI](https://github.com/openai/codex). Where no second vendor is available or approved, the review runs in a fresh same-vendor context and says so in the review document.
  53. ## Workplace use
  54. On an employer's machine, the employer's policies come first. Check extensions against the allowlist, use only approved model vendors, and ask before running `--patch`, which edits an installed extension.
  55. ## Licence
  56. MIT, see [LICENSE](LICENSE). `patterns/PatternsLibrary.md` restates patterns from published catalogs and keeps its own CC BY 4.0 notice and source attributions.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.