convez.ai
Agente
paruff

uFawkesAI

paruff/uFawkesAI

Ver no GitHub
BomPluginScore · de 0 a 100

An AI agent starter template that implements the DORA AI Capabilities Model — built for GitHub Copilot, Claude Code, Cursor, and Codex

Instalação

$npx convez-ai paruff/uFawkesAI

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
0
Acompanhando
1
Saúde
85%
Último commit
há 2 dias
Versão
v2.2.0
Issues abertas
8
Contribuidores
3

README

A tradução para o português deste README ainda está na fila.

uFawkesAI

Configure your AI agent to work the way DORA research proves high-performing teams work. One template. Every agent. DORA AI Capabilities built in from line one. DORA AI Capabilities MIT License uFawkes Family Works with Copilot Works with Claude Code Works with Cursor Open in GitHub Codespaces

The problem

AI agents start every session blank. They do not know your conventions, architecture, or the research-backed practices that improve delivery performance. This template fixes both problems simultaneously.

What you get

  • AGENTS.md — universal agent instruction file, auto-loaded by Copilot, Claude Code, Cursor, Codex, Gemini CLI, Windsurf, and Devin
  • CLAUDE.md symlink — so Claude Code gets the same instructions without a separate file
  • .github/copilot-instructions.md symlink — Copilot's preferred path
  • .github/agents/ — 4 specialist agent profiles: docs, test, review, security
  • .github/instructions/ — scoped instruction files for feature and test work
  • docs/PROMPT_LIBRARY.md — tested prompts for every repeating task, versioned
  • docs/GOLDEN_PATH.md — 10-step idea→deploy workflow
  • docs/TEAM_ARCHETYPE.md — DORA 2025 seven team archetype self-assessment
  • CI workflow with PR size blocking (400 lines), doc freshness monitoring
  • scripts/weekly-metrics.sh — rework rate, PR revision rate, CI cycle time in one view

June 2026: Built for the New Token Billing Model

GitHub Copilot moved to token-based billing on June 1, 2026. This template is pre-optimized for the new cost model:

FileWhat it saves
AGENTS.md (88 lines)~61% always-on context vs. typical 226-line files
.github/skills/On-demand only — zero cost until referenced
.copilotignoreExcludes lock files, build artifacts, generated code
scripts/token-audit.shShows your token footprint before the bill arrives
docs/COPILOT_COST_GUIDE.mdBilling model explained for developers
docs/MODEL_ROUTING_GUIDE.mdWhich mode and model saves the most

Run npm run token-audit after setup to see your baseline.

Copilot Billing Retrofit Prompts

Two ready-to-use prompts let you apply uFawkesAI's token optimization to any existing repo — not just new projects.

PromptToolWhat it producesCost
docs/COPILOT_BILLING_HANDOFF_PROMPT_LOCAL.mdOllama + Gemma 4 E4BAGENTS.md, 4 skill files, .copilotignore, 2 cost guidesFree
docs/COPILOT_BILLING_HANDOFF_PROMPT_SCRIPTS.mdCopilot / Claude Codetoken-audit.sh, setup.sh hardening, CI placeholder check~$0.50

Run the local prompt first (zero cost), then the scripts prompt for the bash work. Together they take about 90 minutes and reduce always-on context by 40–60%.

See docs/COPILOT_BILLING_HANDOFF_PROMPT_LOCAL.md to get started.

5-minute quick start

# 1. Use this template (click "Use this template" on GitHub) or clone
git clone https://github.com/paruff/uFawkesAI.git my-project
cd my-project

# 2. Run setup (creates symlinks, installs hooks)
./scripts/setup.sh

# 3. Customise for your project (replace all [PLACEHOLDER] sections)
# Start with AGENTS.md — everything else derives from it
code AGENTS.md

# 4. Assign your first issue to Copilot or Claude Code
# Your agent now knows your project from session one

Writing your first agent-ready issue

Use the Feature — Assign to Agent issue template (.github/ISSUE_TEMPLATE/feature.yml). Here is an example of a well-formed feature issue that Copilot can implement immediately:


Title: [FEAT] Add weekly rework rate summary to docs/METRICS.md

User Story:

As a team lead, I want to run npm run metrics and see my rework rate for the past 14 days, so that I know immediately whether AI output quality is improving or degrading.

Acceptance Criteria:

  • AC1: npm run metrics outputs a rework rate percentage calculated from the last 14 days of git history
  • AC2: The output is appended as a dated snapshot block in docs/METRICS.md
  • AC3: If rework rate exceeds 20%, the script prints a warning: ⚠ Rework rate > 20% — stop features, fix AGENTS.md first

DORA AI Capability: Rework rate / DORA metrics

Context Files to Read:

  • scripts/weekly-metrics.sh — existing metrics script to extend
  • docs/METRICS.md — target file for snapshot output
  • AGENTS.md §9 — rework rate definition and target thresholds

Constraints and Out of Scope:

  • Do not modify existing scripts/setup.sh
  • Do not add new npm dependencies

Definition of Done:

  • All acceptance criteria met
  • Failing tests written before implementation
  • npm run preflight passes
  • AI-Assisted Review Block completed in PR

Assign to: Copilot ✓


What makes this issue agent-ready?

  • Acceptance criteria are testable (each maps to a specific assertion)
  • Context files are listed so the agent reads the right code first
  • Out-of-scope constraints are explicit (preventing unwanted changes)
  • The Definition of Done gives the agent a self-check before opening the PR

DORA AI Capabilities implemented

DORA AI CapabilityFile(s) in this template
Clear and communicated AI stanceAGENTS.md §1, docs/AI_POLICY.md
Healthy data ecosystemsdocs/CHANGE_IMPACT_MAP.md, docs/API_SURFACE.md
AI-accessible internal dataAGENTS.md §3 context index, .vscode/settings.json
Strong version control practicesCI PR size block, conventional commits standard
Working in small batches400-line PR limit, TDD requirement, feature flags
User-centric focusdocs/GOLDEN_PATH.md, docs/VALUE_STREAM_MAP.md
Quality internal platformsdocs/GOLDEN_PATH.md, agent specialist profiles

Works with

  • GitHub Copilot (native AGENTS.md support since Aug 2025 — server-side auto-load)
  • Claude Code (reads AGENTS.md as fallback; CLAUDE.md symlink for direct load)
  • Cursor (root AGENTS.md auto-detected; .cursor/rules/ for scoped instructions)
  • OpenAI Codex (AGENTS.md is the native format)
  • Gemini CLI (configure via .gemini/settings.json)
  • Windsurf, Devin, Aider — via AGENTS.md

Part of the uFawkes family

uFawkesAI is the AI plane in a five-stack delivery system: it sets policy, context, guardrails, and workflows so every other stack can move faster with less rework.

Integration entry point: docs/UFAWKES_INTEGRATION.md

StackRoleLink
uFawkesAIAI plane (agent policy, context, controls)paruff/uFawkesAI
uFawkesPipeCI/CD and delivery pipeline planeparuff/uFawkesPipe
uFawkesObsObservability and reliability planeparuff/uFawkesObs
uFawkesAppProduct application planeuFawkesApp (planned)
uFawkesDataData and analytics planeuFawkesData (planned)

Quick integration hooks:

  • uFawkesObs: set OTEL_EXPORTER_OTLP_ENDPOINT and OTEL_SERVICE_NAME in the instrumented runtime/service that uses this template; this repository does not emit OTEL spans by itself.
  • uFawkesDORA: npm run metrics currently summarizes local git/coverage signals; GITHUB_TOKEN, GITHUB_OWNER, and GITHUB_REPO are for external/future GitHub API-backed DORA collectors.
  • uFawkesPipe: run the Golden Path (docs/GOLDEN_PATH.md) so AI-authored PRs flow through the deliveryd CI contract (uFawkesPipe pipeline contract: paruff/uFawkesPipe).

Rework rate

DORA 2025 added rework rate as a new core metric. It is the earliest signal that AI output quality is degrading. This template's scripts/weekly-metrics.sh calculates it from your git history. Target: < 10%. Above 20%: stop features, fix AGENTS.md first.

Contributing

See CONTRIBUTING.md (planned in AI-006).

License

MIT

Ver o original no GitHub

Bom para

Relacionados