本文へ移動
cccskills
無料GitHub で公開

create-issue

Craft a single, high-quality GitHub issue from a developer's description of a bug, unexpected behavior, or desired enhancement. Walks through reproduction, codebase exploration, and structured drafting so the issue is actionable on arrival. Use when user wants to file an issue, report a bug, request a feature, or document a problem they've encountered.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md6.8 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Create Issue

Turn a developer's rough description of a problem or idea into a single, well-structured GitHub issue that is actionable the moment someone reads it.

This skill produces one issue at a time.

Principles

Actionable on arrival

A good issue lets someone start working without asking follow-up questions. Every issue must answer: what happened (or what should happen), how to see it, and how to know it's fixed.

Behavioral, not procedural

Describe what the system does vs what it should do. Avoid file paths and line numbers — they go stale within hours. Name types, interfaces, and behavioral contracts instead.

Honest about unknowns

If reproduction is intermittent, say so. If the root cause is unclear, say so. An honest "cause unknown, here's what I've ruled out" is more useful than a guess that sends someone down the wrong path.

Process

1. Understand the problem

Work from whatever the developer tells you. Ask clarifying questions one at a time — don't dump a questionnaire. Focus on:

  • What happened — the actual behavior they observed
  • What they expected — the behavior they wanted
  • When it started — did it work before? What changed?
  • How often — every time, sometimes, once?

If the developer passes an issue reference, error message, stack trace, or log output, use that as your starting point instead of asking them to re-describe it.

Stop asking when you have enough to write a clear description. Two or three targeted questions is usually enough — don't grill beyond what the issue needs.

2. Explore the codebase

If you have access to the codebase, explore the area relevant to the problem. Look for:

  • The domain vocabulary used in this area (use the project's CONTEXT.md glossary if one exists)
  • Existing tests that cover the behavior
  • Related issues or TODOs in the code
  • ADRs that explain decisions in this area

Use what you find to write the issue in the project's own language, not the developer's paraphrase.

3. Attempt reproduction (bugs only)

For bug reports, trace the relevant code path and run existing tests if possible. Do not claim a reproduction you didn't actually observe — an honest "could not reproduce" is more useful than a fabricated one.

Report what you find:

  • Reproduced — describe the exact steps and what you observed
  • Partially reproduced — you saw related behavior but not the exact symptom
  • Could not reproduce — describe what you tried

A confirmed reproduction with a code path makes a dramatically stronger issue. But a failed reproduction is still valuable — note it in the issue so the next person doesn't repeat your work.

4. Classify

Determine the issue type:

  • Bug — something is broken or behaving contrary to its documented/intended behavior
  • Enhancement — new capability or improvement to existing behavior

A thing that "works" but produces wrong results is a bug, not an enhancement. If the distinction is genuinely unclear from the description alone, ask rather than guess — misfiling an enhancement as a bug skews labeling and metrics.

5. Draft the issue

Write the issue using the template below. Present the draft to the developer and ask:

  • Does this accurately describe the problem?
  • Is anything missing or overstated?

Iterate until the developer approves.

6. Publish

Attempt to post the approved issue to the issue tracker using an available tool (e.g. gh CLI or GitHub MCP). Apply the agreed labels. Report back the issue number and URL.

If no publishing tool is available, output the final issue body as a fenced markdown block so the developer can paste it manually:

[issue body here]

Issue body template

Bug

Summary

One or two sentences: what is broken, and what is the impact.

Current behavior

What the system does now. Be specific — include error messages, wrong outputs, or unexpected states. Paste stack traces or logs in a code block if available.

Expected behavior

What the system should do instead. Reference documented behavior, type contracts, or prior working behavior if applicable.

Steps to reproduce

  1. Step one
  2. Step two
  3. Step three

Environment (include only fields relevant to this bug)

  • Runtime:
  • OS:
  • Browser / client:
  • Relevant config:

Frequency: Every time / Intermittent (~X% of attempts) / Once

Investigation notes

What has been explored so far: code paths traced, hypotheses tested, things ruled out. Omit this section if no investigation has been done.

Acceptance criteria

  • Criterion 1
  • Criterion 2

Enhancement

Summary

One or two sentences: what capability is missing or what improvement is proposed.

Motivation

Why this matters. Describe the user problem or workflow gap — not the solution. If this came from a real scenario, describe it.

Proposed behavior

What the system should do. Describe the end-to-end behavior from the user's perspective. Mention key interfaces or types that would be involved if they already exist in the codebase.

Alternatives considered

Other approaches and why they are less suitable. Omit if there is genuinely only one reasonable approach.

Open questions

Design decisions or unknowns that should be resolved before implementation begins. Omit if none.

Acceptance criteria

  • Criterion 1
  • Criterion 2
  • Criterion 3

Operating rules

  • Ask clarifying questions one at a time — never more than one per message
  • Do not skip the draft-and-review loop even if the developer says "just file it"
  • Never publish without explicit developer approval of the draft
  • Do not add file paths or line numbers to the issue body
  • If you cannot reproduce a bug, say so — do not fabricate or imply a reproduction
  • Output a copyable markdown block any time publishing fails or no tool is available

What makes a bad issue

Avoid these — they waste everyone's time:

  • Vague symptoms: "it's broken" / "doesn't work" / "something is wrong with X"
  • Solution-first: jumps to "add a flag" or "change the config" without stating the problem
  • Stale references: file paths, line numbers, or commit SHAs that will drift within days
  • Missing reproduction: "I saw an error" with no steps, environment, or frequency
  • Kitchen-sink scope: five different problems bundled into one issue

If the developer's description falls into any of these, the grilling in Step 1 should fix it before the issue is written.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

adhd

無料

Formats output for ADHD-friendly action — lead with the next step, number tasks, restate state, cut fluff. Use when named (adhd), when the user mentions ADHD or focus/executive-function constraints, or when they ask for steps they can execute immediately.

日本語の概要は準備中です。原文の説明を表示しています。

nateslabach/skills62026年9月29日 更新

Devil's-advocate attack on AI workflow designs — threat model, assumption challenges, code critiques, and hypothetical redesigns with no file edits. Use for "attack this design", devil's advocate, or hostile challenge on agents/tools/summarization/tokens. Not for PR peer review (code-review) or self-review that applies fixes (ai-systems-review). Prefer: this skill → decide → ai-systems-review.

日本語の概要は準備中です。原文の説明を表示しています。

nateslabach/skills62026年9月29日 更新

Hostile self-review of recently written AI workflow code that then applies fixes (agents, tools, chat context, summarization, tokens, attachments). Use after generating AI systems code, or when asked to self-review and harden it. Not for PR peer review (code-review) or no-edit devil's-advocate attacks (adversarial-challenge). Prefer: adversarial-challenge → decide → this skill.

日本語の概要は準備中です。原文の説明を表示しています。

nateslabach/skills62026年9月29日 更新

Writes in ASD-STE100 simplified technical English — short sentences, plain words, active voice. Use when the user wants terse technical prose, names this skill, or asks for STE100.

日本語の概要は準備中です。原文の説明を表示しています。

nateslabach/skills62026年9月29日 更新

branding

無料

Builds personal brand assets using Seth Godin's frameworks: signature symbol, slogan, surprise (Purple Cow), salient idea, and signature story. Use when named (branding), or when the user asks for a brand symbol, slogan, tagline, Purple Cow, counterintuitive angle, core idea, brand story, or personal brand positioning.

日本語の概要は準備中です。原文の説明を表示しています。

nateslabach/skills62026年9月29日 更新

Disables the AI's default hedging, diplomatic softening, and social-nicety layers to deliver cold, unfiltered, consequentialist analysis on any topic. Use when the user wants the AI's actual assessment without padding, regardless of how harsh or uncomfortable the honest answer is.

日本語の概要は準備中です。原文の説明を表示しています。

nateslabach/skills62026年9月29日 更新

nateslabach のスキルをすべて見る

このスキルの問題を報告する