agentleFS
mckinseymckinsey/agents-at-scale-ark/.claude/skills/issue-creation/SKILL.md

issue-creation

Skill439 starsChanged 7 months agoApache-2.0View on GitHub

Structured workflow for drafting NEW GitHub issues with codebase research, duplicate detection, and testing approach. Always asks clarifying questions and shows the draft for approval before creating. For searching, listing, viewing, or updating existing issues, use the "issues" skill instead.

7 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use chained-pr, work-unit-commits and 3 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • joeldevzSkill

    chained-pr

    Split large changes (>400 lines OR >5 slices) into a series of reviewable PRs. Two strategies — stacked-to-main and feature-branch-chain. Prevents the 100-conflict merge disaster.

    Also in 3 projects that use this

  • joeldevzSkill

    work-unit-commits

    Commit discipline. Each commit is a self-contained, reviewable work unit — code + tests + docs together. Stops the 190-commit PR disaster before it starts.

    Also in 3 projects that use this

  • alex-ilgayevSkill

    go-testing

    Handles all Golang testing tasks including running tests, writing new tests, and fixing test failures. Follows MCPSpy testing conventions with require for critical assertions and assert for non-critical ones.

    Also in 3 projects that use this

  • +2 more from joeldevz

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. 0 are common, 16 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 16

    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:

    Create well-researched, problem-focused GitHub issues for mckinsey/agents-at-scale-ark.

    line 8
  2. Distinctive:

    Use when the user asks to create an issue, report a bug, request a feature, or track work.

    line 12
  3. Distinctive:

    Follow these steps in order.

    line 16
  4. Distinctive:

    ALWAYS ask clarifying questions before doing any research or drafting.

    line 20
  5. Distinctive:

    Do NOT proceed to Step 2 until the user has answered your questions.

    line 29
  6. Distinctive:

    Before writing anything, investigate the relevant code:

    line 33
  7. Distinctive:

    Use Grep, Glob, Read, and the Explore agent as needed.

    line 39
  8. Distinctive:

    Search existing issues thoroughly in the main repo:

    line 45
  9. Distinctive:

    If a duplicate exists in either repo, tell the user and link to it instead of creating a new issue.

    line 63
  10. Distinctive:

    If related issues exist in either repo, note them as dependencies or references in the new issue.

    line 64
  11. Distinctive:

    ALWAYS present the full draft (title + body) to the user before creating the issue.

    line 96
  12. Distinctive:

    Do NOT create the issue until the user approves.

    line 100
  13. Distinctive:

    Only after the user approves the draft:

    line 104
  14. Distinctive:

    Always ask clarifying questions first: Never skip straight to research or drafting.

    line 120
  15. Distinctive:

    Research first: Every issue must include codebase research findings.

    line 124
  16. Distinctive:

    No duplicates: Always check existing issues before creating.

    line 125
  17. Distinctive:

    Always add "needs grooming" label: Every issue created by this skill gets this label.

    line 127

“Common” is measured: kept by at least 6 in 10 of the 7 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

Issue Creation

Create well-researched, problem-focused GitHub issues for mckinsey/agents-at-scale-ark.

When to use this skill

Use when the user asks to create an issue, report a bug, request a feature, or track work. This skill enforces a research-first approach before creating any issue.

Process

Follow these steps in order. Do NOT skip steps.

Step 1: Ask clarifying questions

ALWAYS ask clarifying questions before doing any research or drafting. Never assume you have enough context. Use AskUserQuestion to gather:

  • What exactly is the problem? (Get specifics, not just a vague description)
  • Who is affected and how severely?
  • How is the problem reproduced? (for bugs)
  • What is the motivation or trigger? (for features)
  • Is there any urgency or deadline?
  • Are there any constraints the user already knows about?

Do NOT proceed to Step 2 until the user has answered your questions. If the user's initial description is detailed, still confirm your understanding by summarizing back and asking if anything is missing.

Step 2: Research the codebase

Before writing anything, investigate the relevant code:

  • Find the files, modules, and components involved
  • Understand the current behavior and architecture
  • Identify the scope and blast radius of the problem
  • Note relevant file paths and code references

Use Grep, Glob, Read, and the Explore agent as needed. Include key findings in the issue's Context section so reviewers can orient themselves without re-doing the research.

Important: The purpose of research is to understand the problem's scope and surface area — NOT to prescribe a solution. Do not let research findings leak into prescriptive implementation steps. Knowing which files are involved helps the implementer orient; telling them what to change in those files anchors them on a path that may be wrong.

Step 3: Check for duplicates and dependencies

Search existing issues thoroughly in the main repo:

gh search issues --repo mckinsey/agents-at-scale-ark "<keywords>" --json number,title,state,labels --jq '.[] | "\(.number) [\(.state)] \(.title)"'

gh issue list --repo mckinsey/agents-at-scale-ark --state open --json number,title,labels --jq '.[] | "\(.number) \(.title)"' | grep -i "<keyword>"

If the issue relates to the marketplace (observability, community services, Phoenix, Langfuse, optional service integrations, or anything in the marketplace repo), also search and research mckinsey/agents-at-scale-marketplace:

gh search issues --repo mckinsey/agents-at-scale-marketplace "<keywords>" --json number,title,state,labels --jq '.[] | "\(.number) [\(.state)] \(.title)"'

gh issue list --repo mckinsey/agents-at-scale-marketplace --state open --json number,title,labels --jq '.[] | "\(.number) \(.title)"' | grep -i "<keyword>"

Also research the marketplace codebase for relevant patterns, existing implementations, and architectural context using the gh CLI or web fetch against https://github.com/mckinsey/agents-at-scale-marketplace. Include marketplace findings in the issue Context section when relevant.

  • If a duplicate exists in either repo, tell the user and link to it instead of creating a new issue.
  • If related issues exist in either repo, note them as dependencies or references in the new issue.

Step 4: Draft the issue

Use this template:

## Problem

[Clear description of what is broken, missing, or suboptimal. Focus on the problem, NOT the solution.]

## Context

[Codebase research findings: relevant files, current behavior, architecture context. Include file paths.]

## Impact

[Who is affected? What breaks or degrades? How severe is this?]

## Related Issues

- #N — [brief description of relationship]

## Testing Approach

- [How to verify the problem is fixed]
- [What test types are appropriate: unit, integration, e2e]
- [Edge cases to cover]

Step 5: Show the draft and ask for edits

ALWAYS present the full draft (title + body) to the user before creating the issue. Use AskUserQuestion to ask:

  • Does the draft look good?
  • Any sections to add, remove, or reword?

Do NOT create the issue until the user approves. If they request edits, apply them and show the updated draft again.

Step 6: Create the issue

Only after the user approves the draft:

gh issue create --repo mckinsey/agents-at-scale-ark \
  --title "<type>: <concise problem description>" \
  --body "$(cat <<'EOF'
<issue body from step 5>
EOF
)" \
  --label "needs grooming"

Title must use conventional commit prefix: feat:, fix:, docs:, chore:, refactor:, etc.

Rules

  1. Always ask clarifying questions first. Never skip straight to research or drafting. Confirm understanding before proceeding.
  2. Focus on the problem, not the solution. The issue author's job is to be the expert on the problem. Never propose a design, implementation approach, or specific code changes unless the user explicitly asks for one.
  3. Research informs scope, not implementation. Codebase research belongs in the Context section to help the implementer orient. It must NOT become prescriptive implementation steps. Knowing which files are involved is useful context; telling the implementer what to change in those files anchors them on a path that may be wrong.
  4. No uninformed specificity. Implementation details written by someone without deep codebase knowledge create false confidence and anchoring bias. A vague-but-accurate issue is more useful than a specific-but-wrong one.
  5. Research first. Every issue must include codebase research findings.
  6. No duplicates. Always check existing issues before creating.
  7. Testing approach required. Suggest how to verify the fix/feature.
  8. Always add "needs grooming" label. Every issue created by this skill gets this label.
  9. Link dependencies. Reference related issues when they exist.

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.