creating-instructions
Skill118 starsChanged 2 months agoMITView on GitHub
Creates GitHub Copilot instruction files for VS Code, including repository guidance and scoped `.instructions.md` rules. Use when encoding project conventions, choosing `applyTo` patterns, or shipping install-ready instruction examples with rationale and guardrails.
We split this skill into the rules it gives an agent
16 rules. About 0 are common, 12 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈12
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Use this skill when the goal is to package stable rules for GitHub Copilot instead of writing one-off guidance in chat.
- Distinctive: line 15
Use this skill when the user asks for things like:
- Distinctive: line 38
no workflow logic, no agent personality, and no prompt-specific scripting
- Distinctive: line 42
Rules over essays: - prefer short, stable guidance that survives multiple tasks.
- Distinctive: line 44
Reason when needed: - if a rule is surprising, explain the consequence in the same bullet.
- Distinctive: line 46
Context is expensive: - avoid loading broad instructions that repeat what tools already enforce.
- Distinctive: line 69
Use this decision rule:
- Very specific: line 71
create copilot-instructions.md when the rules apply to nearly every task in the repository
- Very specific: line 72
create *.instructions.md when the rules apply to a clear file pattern, directory, or workflow boundary
- Very specific: line 80
write frontmatter with description, applyTo, title, and name when helpful
- Distinctive: line 81
keep each rule compact and testable
- Distinctive: line 82
add examples only for non-obvious conventions
- Distinctive: line 88
keep it opinionated enough to be useful but not so exhaustive that it becomes background noise
- Distinctive: line 99
Use these supporting resources:
- Very specific: line 108
Check the final result against ./resources/instructions-quality-checklist.md.
- Distinctive: line 115
no duplication of agent behavior, prompt workflow, or skill instructions
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Creating Copilot Instructions
Use this skill when the goal is to package stable rules for GitHub Copilot instead of writing one-off guidance in chat. It helps turn repository conventions into clean instruction files that are easy to install, reason about, and maintain.
When to Use
Use this skill when the user asks for things like:
- "create a repo-wide Copilot instruction file"
- "write a
.instructions.mdfile for these tests" - "choose the right
applyTopattern" - "turn our coding rules into installable instructions"
- "fix an instruction file that is too broad or too vague"
Typical scenarios:
- first-time setup of
copilot-instructions.md - adding targeted guidance for a folder, language, or test type
- splitting an overgrown instruction file into smaller units
- preparing public examples for a team repository or a shared collection
What Good Looks Like
A strong instruction contribution usually includes:
- the right instruction type for the job: repo-wide or scoped
- a narrow, predictable scope
- short declarative rules with a reason when the rule is not obvious
- examples only where they materially improve compliance
- no workflow logic, no agent personality, and no prompt-specific scripting
Instruction Design Rules
- Rules over essays - prefer short, stable guidance that survives multiple tasks.
- Scope first - the
applyTopattern is part of the design, not an afterthought. - Reason when needed - if a rule is surprising, explain the consequence in the same bullet.
- Examples beat abstraction - include a preferred and avoided example for tricky patterns.
- Context is expensive - avoid loading broad instructions that repeat what tools already enforce.
Workflow
Phase 0: Frame the request
Clarify or infer:
- whether the user needs repo-wide guidance or scoped instructions
- who will install the file: one user, a team, or a public collection
- which files or folders must be affected
- what conventions are stable enough to encode as rules
Phase 1: Study the source material
Before drafting:
- inspect existing instructions, README notes, and representative source files
- separate lint or formatter-enforced behavior from guidance that still needs human judgment
- collect only durable conventions; avoid temporary migration notes unless explicitly requested
Phase 2: Choose the file shape
Use this decision rule:
- create
copilot-instructions.mdwhen the rules apply to nearly every task in the repository - create
*.instructions.mdwhen the rules apply to a clear file pattern, directory, or workflow boundary - split files when one audience would otherwise inherit unrelated rules
Phase 3: Draft the instruction file
For scoped instruction files:
- choose a descriptive filename in kebab case
- write frontmatter with
description,applyTo,title, andnamewhen helpful - keep each rule compact and testable
- add examples only for non-obvious conventions
For repo-wide instruction files:
- describe architecture, stack, folder responsibilities, commands, and non-obvious conventions
- organize by headings so Copilot can find the right rule quickly
- keep it opinionated enough to be useful but not so exhaustive that it becomes background noise
Phase 4: Create the artifact bundle
Always prepare:
- the main instruction file
- a short installation note or target path
- at least one reference template or example when the instruction is meant to be reused
- a brief summary of why the scope and file type were chosen
Use these supporting resources:
./resources/copilot-instructions.template.md./resources/custom-instructions.template.instructions.md./resources/copilot-instructions.example.md./resources/custom-instructions.example.instructions.md
Phase 5: Validate before handoff
Check the final result against ./resources/instructions-quality-checklist.md.
Pay extra attention to:
applyTobeing narrow enough to avoid accidental always-on context- rules being declarative, not step-by-step workflows
- examples matching the claimed rule
- no duplication of agent behavior, prompt workflow, or skill instructions
Common Failure Modes
- using
applyTo: '**'for advice that only matters in one part of the repo - turning the file into a tutorial instead of a rule set
- copying linter or formatter defaults instead of encoding missing conventions
- mixing multiple unrelated audiences into one instruction file
- writing rules without enough context for Copilot to understand why they matter
Resource Map
./resources/copilot-instructions.template.md- repo-wide starting point for a project constitution./resources/custom-instructions.template.instructions.md- scoped instruction template with frontmatter./resources/copilot-instructions.example.md- worked repo-level example for a test automation repository./resources/custom-instructions.example.instructions.md- worked scoped example with concrete rules./resources/instructions-quality-checklist.md- final review checklist before shipping
Definition of Done
A task using this skill is complete when:
- the chosen instruction type matches the actual scope
- the file is install-ready and placed in the right target location
- the rules are concise, durable, and non-duplicative
- supporting template or example material exists when the artifact is meant for reuse
- the final result passes the quality checklist without obvious ambiguity
In this repository51 other files it gives its agents.
- analyzing-quality-metricsSkill
- analyzing-regression-scopeSkill
- api-playwright-test-developerSkill
- assessing-comprehension-debtSkill
- assessing-release-readinessSkill
- attesting-manual-verificationSkill
- auditing-accessibilitySkill
- code-review-advancedSkill
- code-reviewSkill
- communicating-quality-findingsSkill
- creating-custom-agentsSkill
- creating-hooksSkill
- creating-orchestration-packsSkill
- creating-pluginsSkill
- creating-promptsSkill
- creating-skillsSkill
- deriving-a-quality-contractSkill
- designing-functional-testsSkill
- designing-test-dataSkill
- documenting-test-suitesSkill
- generating-quality-gate-workflowsSkill
- governing-quality-waiversSkill
- handling-sensitive-test-dataSkill
- migrating-tests-to-playwrightSkill
- mocking-network-and-timeSkill
- planning-exploratory-testingSkill
- prd-generatorSkill
- recording-change-intentSkill
- reporting-bugsSkill
- requirements-test-coverage-mapperSkill
- reviewing-ai-output-groundednessSkill
- running-visual-regression-testsSkill
- scoping-change-relevanceSkill
- static-code-analysis-typescriptSkill
- tech-debt-analysisSkill
- test-driven-development-quickSkill
- test-driven-developmentSkill
- testing-agent-skillsSkill
- testing-api-contractsSkill
- testing-application-securitySkill
- testing-llm-featuresSkill
- testing-llm-guardrailsSkill
- testing-performance-and-loadSkill
- tracing-requirements-to-codeSkill
- tracking-quality-trendsSkill
- unslop-answersSkill
- unslop-testsSkill
- verifying-acceptance-criteriaSkill
- verifying-change-coverageSkill
- writing-unit-tests-quickSkill
- writing-unit-testsSkill
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
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.

