| name | github-issue-implementer |
| description | Use when the user asks to read a GitHub issue, understand requirements, implement the fix or feature in code, add tests, and summarize changes with verification steps. Trigger phrases: GitHub issue, issue URL, implement issue, fix issue #123, resolve ticket. |
| user-invocable | true |
GitHub Issue Implementer
Language / Sprache
- Provide all user-facing outputs in German and English.
- Gib alle nutzerseitigen Ausgaben auf Deutsch und Englisch aus.
- For each major section in summaries, present German first, then English.
- Stelle in jedem Hauptabschnitt von Zusammenfassungen zuerst Deutsch und danach Englisch dar.
- Keep issue terminology aligned across both languages (for example requirement/Anforderung, blocker/Blocker).
- Halte die Issue-Terminologie in beiden Sprachen konsistent (z. B. requirement/Anforderung, blocker/Blocker).
Goal
Read a GitHub issue, extract actionable requirements, implement the requested behavior in the current workspace, verify with tests or checks, and return a concise implementation summary.
Inputs You Need
- Issue reference: full URL (preferred) or
owner/repo#number
- Target branch or constraints (if provided)
- Acceptance criteria (if provided)
- Any environment limits (offline, no credentials, no network)
If key inputs are missing, ask only the minimum clarifying questions required to proceed.
Prerequisite: Review Project Documentation First
Before mapping requirements or editing code, review project guidance and conventions in:
AGENTS.md (if present)
README.md (if present)
/docs/* files (if present)
Use these sources to align implementation approach, coding conventions, test strategy, and validation commands with the project's documented standards.
Orchestrator Sequence (Main Skill + Subagents)
Use this mandatory execution order:
- Main skill (
github-issue-implementer): parse issue, derive requirements, and map impacted code locations.
- Subagent skill (
github-issue-implementation-executor): implement requirements and update tests.
- Subagent skill (
github-issue-validation-runner): run prioritized validation checks and report evidence.
- Main skill (
github-issue-implementer): consolidate outputs and produce the final bilingual summary.
Handoff requirements between steps:
- Main -> Implementation subagent:
- Requirements list, acceptance criteria, impacted files/symbols, constraints, assumptions.
- Implementation -> Validation subagent:
- Changed file list, test changes, implementation assumptions, known risk areas.
- Validation -> Main:
- Commands executed, pass/fail/skip results, diagnostics, residual risks, skipped checks.
If a subagent returns blockers, stop the sequence at that step, ask only blocking clarifications, then resume from the blocked step.
Workflow
- Parse and fetch the issue
- If a URL is provided, read it with available web tools.
- If
owner/repo#number is provided, build the issue URL and fetch it.
- Collect: title, body, labels, linked tasks, and relevant comments.
- If the issue cannot be fetched (private repo, auth, network), ask the user to paste the issue content and continue.
- Convert issue text into implementation requirements
- Produce a short requirements list.
- Separate must-have behavior from optional improvements.
- Identify acceptance criteria and edge cases.
- Note ambiguities and resolve only blocking ones with targeted questions.
- Map requirements to codebase locations
- Search the workspace for entry points, related symbols, tests, and configs.
- Prefer existing architecture and conventions over introducing new patterns.
- Identify the smallest safe set of files to change.
- Implement changes end-to-end
- Delegate implementation to a dedicated subagent using the separate skill
github-issue-implementation-executor.
- Skill file:
.github/skills/github-issue-implementation-executor/SKILL.md
- Subagent inputs: requirements, affected files/symbols, coding constraints, and acceptance criteria.
- Require the subagent to return:
- List of changed files and rationale.
- Summary of implementation decisions and assumptions.
- Any blockers requiring user input.
- Validate
- Delegate validation to a dedicated subagent using the separate skill
github-issue-validation-runner.
- Skill file:
.github/skills/github-issue-validation-runner/SKILL.md
- Subagent inputs: changed files, project toolchain, and required quality checks.
- Require the subagent to return:
- Commands executed.
- Pass/fail results with key diagnostics.
- Remaining risks, skipped checks, and justification.
- Summarize outcome
- What was implemented.
- Files changed and why.
- Verification performed and results.
- Any residual risks or follow-up tasks.
Implementation Rules
- Prefer minimal, targeted diffs.
- Preserve public APIs unless the issue explicitly requires a breaking change.
- Keep naming and style consistent with nearby code.
- Add short comments only where logic is not obvious.
- Never fabricate issue details; clearly mark assumptions.
- If the working tree is dirty, do not revert or overwrite unrelated existing changes; limit edits to issue-scoped files and integrate safely with current workspace state.
Clean Code and Maintainability Requirements
- Write readable code first: choose intention-revealing names for variables, functions, classes, and tests.
- Keep units small and cohesive: each function/class should have one clear responsibility.
- Avoid deep nesting and overly complex conditionals; favor guard clauses and simple control flow.
- Prefer composition and existing abstractions over ad-hoc duplication.
- Do not introduce clever or opaque patterns when a straightforward solution is available.
- Keep side effects explicit and localized; avoid hidden state changes.
- Maintain stable interfaces and contracts unless the issue explicitly requires a change.
- Add or improve tests to cover primary behavior, edge cases, and regressions introduced by the issue.
- Ensure error handling is clear and actionable; fail with meaningful messages where appropriate.
- Leave touched code in a better state: small opportunistic cleanups are allowed if directly adjacent and low risk.
Output Format
Use this structure in the final response:
- Issue Understanding
- Problem statement
- Acceptance criteria
- Changes Implemented
- Validation
- Commands/checks run
- Results
- Notes
- Open questions, limitations, or follow-ups
Failure Handling
- If issue access fails: request pasted issue content and continue.
- If workspace lacks needed code: explain the gap and propose the smallest viable next step.
- If tests are missing: add focused tests where possible and document coverage limits.