retro-try-handler

Detects, classifies, and implements open GitHub retrospective Try issues as pull requests.

Updated Aug 17, 2026
One-click install
npx skills add https://github.com/kai-kou/gem-hunter --skill retro-try-handler-kai-kou
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: retro-try-handler
Source: https://github.com/kai-kou/gem-hunter/tree/main/.claude/skills/retro-try-handler
Command: npx skills add https://github.com/kai-kou/gem-hunter --skill retro-try-handler-kai-kou

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Retrospective meetings generate improvement ideas (Try items) that often sit untouched as open GitHub Issues. This Skill closes that gap by automatically finding unhandled Try issues, implementing the small and medium ones as code or documentation changes, and shipping them through reviewed, auto-merged pull requests. ## Core Features & Use Cases - Automatic Issue Triage: Fetches open type:retro-try and update-type issues via the GitHub MCP API, sorts them by urgency/priority labels, and auto-closes stale items untouched for 30 days. - Categorized Implementation: Classifies each issue (doc, script, validation hook, skill, tool-update, domain, dev-tool) and applies the minimal change, committing one commit per issue on a dedicated branch. - PR Bundling and Auto-Merge: Bundles same-category small fixes into a single PR, runs a self-review layer, and squash-merges after checks pass, then triggers the retrospective skill again. - Use Case: A scheduled routine fires every two hours; when a Try issue labeled status:waiting-claude has aged 8 hours, this Skill picks it up, edits the referenced rule file, opens a PR, and merges it without human intervention. ## Quick Start Ask the AI to process the open retro-try issues and turn them into pull requests.

Frequently Asked Questions about retro-try-handler

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automatically implement GitHub issues from retrospectives?▼

Label improvement issues with type:retro-try and status:waiting-claude, then invoke this Skill. It fetches open issues via the GitHub MCP list_issues tool, sorts them by urgency and priority labels, implements small and medium items, and opens auto-merged pull requests.

How to auto-close stale GitHub issues that were never started?▼

The Skill's Step 1.5 closes any type:retro-try issue with status:waiting-claude that has no updates for 30 days, using state_reason not_planned and a comment explaining the TTL closure. It processes at most 5 issues per run and skips blocker-labeled items.

Can I use gh CLI instead of GitHub MCP for issue operations?▼

Yes, but only in local environments where gh can reach GitHub directly. In cloud environments gh is blocked with 403 errors, so the Skill uses mcp__github__ tools as the primary path, with gh commands documented as local-only alternatives.

Why does the Skill skip large-effort issues instead of implementing them?▼

Large issues receive only an implementation plan comment and a done_type:D-plan label, limited to one per session. This prevents sessions from consuming their processing limit without closing anything, and unchanged large issues eventually exit via the 30-day TTL closure.

What are the limits on how many issues get processed per run?▼

The limit scales with backlog size: 2 issues for 0-9 open, up to 5 for 30 or more. Tool calls are capped at 8 per turn, so larger batches are split across multiple turns with intermediate reports.