make-issues

Convert code review findings into labeled, deduplicated GitHub issues.

Updated Jun 4, 2026
One-click install
npx skills add https://github.com/PeteRichardson/skills --skill make-issues-peterichardson
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: make-issues
Source: https://github.com/PeteRichardson/skills/tree/main/make-issues
Command: npx skills add https://github.com/PeteRichardson/skills --skill make-issues-peterichardson

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Review reports like PROJECT_REVIEW.md or /code-review output list findings, but turning them into tracked GitHub issues by hand is repetitive and produces inconsistent labels and duplicate issues. This Skill automates that conversion with a consistent label vocabulary and fingerprint-based deduplication. ## Core Features & Use Cases - Finding-to-issue conversion: Parses both project-review Findings tables and code-review tiered sections into a common record, then creates GitHub issues with severity, category, effort, and source labels. - Deduplication via fingerprints: Searches existing issues by file path and scores keyword overlap (Jaccard similarity) against a hidden HTML fingerprint marker, pausing for confirmation only on likely duplicates. - Backfill subcommand: Retrofits fingerprint markers onto pre-existing, hand-filed issues via new comments so they participate in future dedup searches. - Use Case: After running /project-review on a repo, invoke /make-issues to file all Medium+ findings as labeled GitHub issues, skipping anything that duplicates an existing issue. ## Quick Start Ask the AI to run /make-issues in your repository to turn the findings in PROJECT_REVIEW.md into labeled, deduplicated GitHub issues.

Frequently Asked Questions about make-issues

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

FAQPage Schema
How do I turn code review findings into GitHub issues?

Run /make-issues with the path to a PROJECT_REVIEW.md or code-review report. It parses each finding, composes an issue with severity and category labels, checks for duplicates, and creates issues via the gh CLI.

How does duplicate detection work when filing GitHub issues?

The skill searches existing issues by file path, then scores Jaccard keyword overlap against a hidden fingerprint marker stored in issue bodies or comments. Matches above roughly 20 percent pause for user confirmation; below that, issues are created immediately.

Can I file issues for only specific findings from a review report?

Yes. Use the -f flag with finding IDs, such as /make-issues -f F3 F5 for project-review reports or ordinal numbers for code-review reports. IDs not found in the source are reported as skipped while the rest of the run continues.

Does make-issues work with both open and closed GitHub issues?

Yes. Dedup searches use --state all so closed issues are included, since a closed fixed issue reappearing as a live finding is meaningful signal. The backfill subcommand also processes closed issues.

What are the limitations of fingerprint-based issue deduplication?

Dedup is best-effort: a file rename breaks the strongest signal and matching falls back to keywords only, so occasional duplicate issues are expected. Category classification is also an LLM judgment call and may occasionally mislabel a finding.