Automates the process of creating and refining project proposal issues.

Install

mkdir -p .claude/skills/create-issue-not-elm && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/12471" && unzip -o skill.zip -d .claude/skills/create-issue-not-elm && rm skill.zip

Installs to .claude/skills/create-issue-not-elm

Activation

This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.

Create a GitHub proposal issue with spec-level brainstorming. Use when the user wants to file a new proposal (enhancement) issue.
129 chars✓ has a “when” trigger
Intermediate

Key capabilities

  • Parse free-form ideas into structured GitHub issue fields
  • Check for duplicate or related GitHub issues
  • Refine problem, solution, and scope through brainstorming
  • Generate a final draft of a GitHub proposal issue
  • Create a GitHub issue with an enhancement label

How it works

The skill processes a user's free-form idea, structures it into a GitHub proposal template, performs duplicate detection, refines the spec through brainstorming, and then creates the issue on GitHub.

Inputs & outputs

You give it
Free-form proposal description in any language
You get back
URL of the created GitHub proposal issue

When to use create-issue

  • File a new proposal
  • Refine project enhancement ideas
  • Check for existing proposals

About this skill

Create Issue

Create a GitHub proposal issue. Takes a free-form idea (any language), structures it into the project's proposal template fields, checks for duplicates, refines the spec via brainstorming, and files the issue on GitHub. All output is in English.

Flow

/create-issue <proposal description in any language>
  |
  v
1. Argument parsing
  |  - Extract proposal content from args
  |  - No args → stop
  |
  v
2. Preflight
  |  - git repo check
  |  - gh auth check
  |  - Failure → abort with message
  |
  v
3. Initial structuring
  |  - Best-effort mapping to 4-5 template fields + title
  |  - All English output
  |
  v
4. Duplicate detection (two-stage)
  |  - Stage 1: keyword search from initial structure
  |  - Stage 2: semantic check on top 5 (if Stage 1 has hits)
  |  - Likely duplicate / related found → present, ask to continue or abort
  |  - Command failure → warn, skip to step 5
  |
  v
5. Spec-level brainstorming (delegate to superpowers:brainstorming)
  |  - Refine problem, solution, scope
  |  - NOT implementation design
  |
  v
6. Final draft (build issue from refined spec) → approve / edit / abort
  |
  v
7. gh issue create (HEREDOC body, enhancement label) → report URL
  |  - Hint: "Run /brainstorm-issue #<number> to start designing a solution."

Argument Parsing

InputAction
Args providedUse as proposal content, any language
(none)Display "Please provide a proposal description." and stop

Preflight

Run checks in order. Abort on first failure:

CheckCommandFailure
Git repositorygit rev-parse --git-dirAbort: "Not a git repository."
gh authenticatedgh auth statusAbort: "Not authenticated. Run gh auth login."

Initial Structuring

Take the user's free-form input (any language) and produce a best-effort English draft.

Title: Generate a concise English title. Priority: meaning preservation > imperative mood > under 80 characters.

Body fields:

FieldRequiredOutput format
## TypeYesExact literal: New feature or Improvement to existing feature
## Problem / Current BehaviorYesFree-form English prose
## Proposed SolutionYesFree-form English prose
## Affected AreaYesMulti-select, slash-separated from: engine / packages / mods / docs / website / ci/build / other
## Alternatives ConsideredNoInclude if user mentioned alternatives. Otherwise omit section entirely

Rules

  • Do NOT include boilerplate checklists (- [ ] I searched existing issues...).
  • Do NOT include HTML comments.
  • If input is too vague to fill a required field meaningfully, produce a best-effort draft. The brainstorming phase (step 5) will refine it.

This initial structure serves two purposes:

  1. Provides keywords for duplicate detection (step 4)
  2. Provides starting material for the brainstorming delegation (step 5)

Duplicate Detection

Stage 1 — Keyword Search

Extract 3-5 key terms from the initial structure (English). Run:

gh issue list --search "<keywords> is:open" --json number,title,url,labels --limit 10
ResultAction
0 hitsSkip Stage 2, proceed to brainstorming (step 5)
1+ hitsProceed to Stage 2
Command failureWarn ("Duplicate check failed, skipping"), proceed to brainstorming (step 5)

Stage 2 — Semantic Check

Take the first 5 results (search-rank order as returned by gh). Fetch each:

gh issue view <number> --json title,body

Classify each via LLM judgment:

  • Likely duplicate — same problem, similar solution
  • Related — overlapping area, different problem or solution
  • Not duplicate — superficial keyword match only
ResultAction
Likely duplicate or related foundPresent matches, ask to continue or abort
Only "not duplicate"Proceed silently
Individual fetch failureSkip that candidate, continue with remaining

Presentation Format

If likely duplicate or related issues are found, present them:

### Potential Duplicates Found

**Likely duplicate:**
- #42 — Add character switcher to tray menu

**Related:**
- #87 — System tray integration improvements

Continue creating this issue? (yes / no)

User declines → stop. User continues → proceed to brainstorming (step 5).

Spec-Level Brainstorming

Delegate to superpowers:brainstorming via the Skill tool. Pass the initial structure as args, prefixed with:

Refine this proposal at the specification level. Focus on clarifying the problem, sharpening the proposed solution, and identifying the correct scope. Do NOT design implementation — this is about what the feature should do, not how to build it. The output should be a refined version of the proposal fields (Type, Problem, Proposed Solution, Affected Area, Alternatives Considered).

The brainstorming skill runs its normal process:

  1. Explore project context
  2. Ask clarifying questions (one at a time)
  3. Propose 2-3 approaches (for the feature itself, not implementation)
  4. Present refined spec, get user approval
  5. Write spec to docs/superpowers/specs/

Scoping Boundary

In scopeOut of scope
What problem does this solve?Which crate implements it?
What should the user experience be?What data structures are needed?
What are the constraints/edge cases?How do we test it?
What alternatives were considered?What's the implementation plan?

After brainstorming completes, resume at step 6 — building the final issue draft from the refined spec.

Final Draft

Build Final Draft

Map the refined spec from brainstorming back to the proposal template fields:

  • Title: Re-generate if the brainstorming significantly changed the scope. Same rules: meaning preservation > imperative mood > under 80 chars.
  • Body: Rebuild the 4-5 fields from the refined spec. Same format rules as Initial Structuring.

Present to User

Show the full issue exactly as it will appear on GitHub:

### Issue Draft

**Title:** Add tray menu for character switching

---

## Type

New feature

## Problem / Current Behavior

Currently users must open the settings UI to switch between VRM characters...

## Proposed Solution

Add a submenu to the system tray icon that lists available characters...

## Affected Area

engine / mods

## Alternatives Considered

A keyboard shortcut was considered but discarded because...
User actionBehavior
ApproveExecute gh issue create (step 7)
Edit requestsApply edits, re-present. Repeat until approved
AbortStop with no side effects

Issue Creation

Execution

Use a HEREDOC for the body to avoid quote/newline escaping issues:

gh issue create --title "<title>" --label enhancement --body "$(cat <<'EOF'
<body>
EOF
)"
ResultAction
SuccessReport issue URL
FailureDisplay error as-is and abort

Post-Creation

Report the created issue URL and hint at next steps:

Created: <url>

Tip: Run /brainstorm-issue #<number> to start designing a solution.

Constraints

  • All output (title + body) MUST be in English regardless of input language.
  • MUST NOT create an issue without explicit user approval of the full draft.
  • MUST NOT proceed past a failed preflight check.
  • The enhancement label is always applied; no additional label selection step.
  • Headings use exact case-sensitive matches compatible with brainstorm-issue's parser: ## Type, ## Problem / Current Behavior, ## Proposed Solution, ## Affected Area, ## Alternatives Considered.
  • When ## Alternatives Considered is omitted, brainstorm-issue handles this gracefully.

When not to use it

  • When not in a git repository
  • When `gh` is not authenticated
  • When the user declines to continue after duplicate detection

Prerequisites

git repositorygh authenticated

Limitations

  • All output must be in English
  • Cannot create an issue without explicit user approval
  • The `enhancement` label is always applied

How it compares

This skill automates the entire process of drafting, refining, and filing a GitHub proposal issue, including duplicate checks and brainstorming, which is more structured than manually creating an issue.

Compared to similar skills

create-issue side by side with the closest alternatives in the catalog.

SkillInstallsUpdatedSafetyDifficulty
create-issue (this skill)04moReviewIntermediate
task-management165moNo flagsBeginner
compact46moReviewBeginner
daily15moNo flagsBeginner

Try saying

Example prompts that trigger this skill in your AI assistant.

Search skills

Search the agent skills registry