A curated, specification-first library of prompts, instruction packs, skills, and Copilot agents that keeps AI helpers aligned with the spec-kit operating model.
This library provides a central source of reusable prompts, instruction packs, skills, and Copilot agents for AI-assisted development workflows. It enables specification-driven development by keeping AI helpers aligned with a shared operating model.
Teams gain consistent, deterministic automation across repositories. Copy-paste reuse makes onboarding faster, and governance gates ensure that specifications, code, and documentation stay synchronised.
Without shared prompt files, AI assistants drift from agreed standards, produce inconsistent outputs, and lack deterministic validation. Teams end up reinventing the same prompts and struggling to maintain alignment across projects.
Prompts, agents, and skills are written directly against the spec-kit constitution. Instruction packs apply deterministic lint, test, and review rules. Every workflow leans on make lint, make test, and explicit governance gates, keeping behaviour measurable and testable.
- Git
- Make (GNU Make 3.82+)
- A text editor (VS Code recommended for Copilot integration)
# Clone the repository
git clone https://github.com/stefaniuk/promptfiles-copilot.git
cd promptfiles
# Configure the development environment
make config
# Verify quality gates work
make lint && make testGet up and running in minutes:
| Step | Action |
|---|---|
| 1 | ✂️ Copy the prompts or instruction packs you need straight into downstream repositories |
| 2 | 📦 Install instructions by copying guidance into .github/instructions so Copilot inherits rules automatically |
| 3 | 🤖 Use prompts under .github/prompts or agents under .github/agents to plan, spec, and review |
| 4 | 🧪 Validate by running make lint and make test whenever you touch source material |
| 5 | 🧠 Document architectural reasoning in docs/adr for future context |
Expected output: make lint and make test complete successfully with exit code 0.
- Specification-first truth — prompts, agents, and skills are written directly against the spec-kit constitution, so code, docs, and governance remain synchronised
- Consistent guardrails — instruction packs apply deterministic lint, test, and review rules across every repo, ensuring Copilot never drifts from agreed standards
- Deterministic automation — every workflow leans on
make lint,make test, and explicit governance gates, keeping behaviour measurable and testable - Copy-ready building blocks — everything is shippable by folder, making large organisations faster to onboard
- Governance gates — explicit checkpoints between specification and implementation
- This library does not implement the underlying spec-kit framework itself
- It does not provide runtime execution environments for prompts
- It is not a replacement for language-specific linters or test frameworks
The spec-kit lifecycle follows a structured flow:
- Discover the right prompt from the library
- Ground it in a specification using agents like
/speckit.specify - Plan the implementation with
/speckit.plan - Generate tasks with
/speckit.tasks - Implement with
/speckit.implement - Review with governance gates (
/review.speckit-documentation,/review.speckit-code,/review.speckit-test) - Automate every validation step with
make lintandmake test
flowchart TD
constitution["/speckit.constitution"] --> specify["/speckit.specify"]
specify --> needClarification{Need clarification?}
specify -.- specifyNote["💡 Example: Transform the product requirements document #file:PRD.md into a formal specification"]
needClarification -- Yes --> clarify["/speckit.clarify"]
clarify --> specify
needClarification -- No --> plan["/speckit.plan"]
plan -.- planNote["💡 Example (run +1): Run again to verify all items on the Plan Completion Checklist are satisfied"]
plan -.- planNote2["💡 Example: Use Python, uv and pytest as the implementation technologies"]
plan --> domainCoverage{Anything missing?}
domainCoverage -- Yes --> checklist["/speckit.checklist"]
checklist --> plan
checklist -.- checklistNote["💡 Example: Create a checklist for building, assembling and testing the deployment artefacts of all the components being implemented"]
domainCoverage -- No --> tasks["/speckit.tasks"]
tasks -.- tasksNote["💡 Example (run +1): Run again to verify all items on the Tasks Completion Checklist items are satisfied"]
tasks --> consistency{Need consistency check?}
consistency -- Yes --> analyze["/speckit.analyze"]
analyze --> tasks
consistency -- No --> reviewDocs["/review.speckit-documentation"]
reviewDocs -.- reviewDocsNote["💡 Example (run+1): Validate #file:deployment.md checklist, confirm each item is documented, apply sensible defaults where missing or request clarification"]
reviewDocs --> implement["/speckit.implement"]
implement -.- implementNote["💡 Example (run N-times): Phase X"]
implement --> reviewCode["/review.speckit-code"]
reviewCode --> reviewTest["/review.speckit-test"]
classDef source fill:#dbeafe,stroke:#1d4ed8,color:#0f172a,stroke-width:1px;
classDef action fill:#d1fae5,stroke:#34d399,color:#064e3b,stroke-width:1px;
classDef question fill:#fef3c7,stroke:#ea580c,color:#78350f,stroke-width:1px;
classDef review fill:#fce7f3,stroke:#db2777,color:#831843,stroke-width:1px;
classDef docReview fill:#e0e7ff,stroke:#6366f1,color:#312e81,stroke-width:1px;
classDef note fill:#f5f5f4,stroke:#a8a29e,color:#57534e,stroke-width:1px,stroke-dasharray:3;
class constitution source;
class specify,plan,tasks,implement action;
class needClarification,domainCoverage,consistency question;
class clarify,checklist,analyze review;
class reviewDocs,reviewCode,reviewTest docReview;
class specifyNote,checklistNote,planNote,planNote2,tasksNote,implementNote,reviewDocsNote note;
No additional configuration is required beyond the initial setup. The library uses convention over configuration with sensible defaults.
make apply dest=/absolute/path/to/targetWhat gets copied?
.github/agents,.github/instructions,.github/prompts,.github/skills.github/copilot-instructions.md.github/pull_request_template.md(only if missing in the target).specify/memory/constitution.md.specify/scripts/bash,.specify/templatesdocs/adr/ADR-nnn_Any_Decision_Record_Template.mddocs/codebase-overview/,docs/prompts/,docs/.gitignoreproject.code-workspace(only if missing in the target)
Next step: Review git status in the target repo, commit, and run
make lint && make test
# Default: scan Copilot prompt files
make count-tokens
# Scan all markdown, sorted by size
make count-tokens args="--all --sort-by tokens"
# Target specific paths
make count-tokens args=".github/instructions .specify"The report shows:
- Tokens — per-file token counts
- No IDs — counts with identifiers like
[ID-<prefix>-NNN]stripped - Usage % — context window usage (200K baseline)
| Gate | Command | Purpose |
|---|---|---|
| 📄 Documentation | /review.speckit-documentation |
Consistency across spec.md, plan.md, tasks.md |
| ✅ Code Compliance | /review.speckit-code |
Reconcile implementation with spec |
| 🧪 Test Quality | /review.speckit-test |
Ensure healthy test pyramid |
| 🧰 Instructions | /enforce.[tech] |
Lint & test at every delivery phase |
Why governance gates matter
- Deterministic flow — each gate blocks the next phase until resolved
- Auditability — checklist evidence for compliance reviews
- Scalability — repeatable tasks across dozens of teams
- Fewer regressions — catch integration issues early
- Better onboarding — contributors understand the lifecycle from tasks.md
| Pack | Description |
|---|---|
| 🤖 .github/agents | Ready-to-run Copilot agents (analyze, checklist, clarify, constitution, implement, plan, specify, tasks, taskstoissues) tuned for spec-kit ceremonies |
| 💬 .github/prompts | Focused prompt files for documentation reviews, governance gates, tests, and refactoring support |
| 📋 .github/instructions | Coding standards and best practice packs scoped by file glob so Copilot always sees the right rules |
| 🧠 .github/skills | Bundled instructions plus helper assets that extend Copilot's capabilities for niche workflows |
| 📝 .specify/templates | Seed specs, plans, and tasks for new features |
| 📄 docs/adr/ADR-nnn_Any_Decision_Record_Template.md | Opinionated ADR template aligned with spec-kit identifiers |
Prompts use a prefix + category + verb convention to keep fuzzy search fast and predictable:
| Prefix | Purpose | Example |
|---|---|---|
speckit. |
Spec-kit lifecycle steps | speckit.plan.prompt.md |
codebase. |
Evidence-first codebase documentation flows | codebase.01-repository-map.prompt.md |
enforce. |
Instruction compliance enforcement | enforce.python.prompt.md |
review. |
Review and audit prompts | review.speckit-code.prompt.md |
util. |
Operational utilities | util.gh-pr-review.prompt.md |
| Resource | Link |
|---|---|
| Custom Prompts | VS Code Docs |
| Custom Instructions | VS Code Docs |
| Custom Agents | VS Code Docs |
| Custom Skills | VS Code Docs |
| Awesome Copilot | GitHub |
We welcome contributions! See contributing.md for the full guide.
git clone https://github.com/stefaniuk/promptfiles-copilot.git
cd promptfiles
make configmake lint # Run linters (file format, markdown format, markdown links)
make test # Run tests- Raise an issue or PR describing your planned changes
- Keep artefacts in sync — specs, plans, tasks, and docs must align
- Run quality gates —
make lint && make testbefore opening a PR - Follow the constitution and NHS Engineering guidance
.github/agents/— Copilot agent definitions for spec-kit ceremonies.github/instructions/— Coding standards by language/framework.github/prompts/— Task-specific prompt files.github/skills/— Bundled capabilities with supporting assets.specify/— Spec-kit templates and project constitutiondocs/adr/— Architecture decision recordsscripts/— Build and utility scripts
📝 New Prompts
architecture-review.prompt— architect for flowmigrate-from-[tech A]-to-[tech B].prompt
🔧 Workflow Prompts
- Release notes — changelog entries grouped by spec identifiers
This project is licensed under the MIT Licence. See LICENCE.md for details.