Skip to main content
skill CX Codex

gh · prathamdby — Claude Code skill: what it does, how to install

The gh skill lets an agent inspect PRs, discussions, and CI, post replies, and handle GitHub commands that need more than a simple call.

MIT 420 installs ★ 6 updated 2026-11-10

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add prathamdby/skills --skill gh -a claude-code
Open on GitHub

What gh does

The skill covers pull request state, checks, files, review threads, and CI runs. It selects a dedicated script for a PR snapshot, thread data, CI failure analysis, or a reply in one conversation. Execution goes through <anchor>/scripts/run rather than calling Node directly.

Requests can target a PR number or URL, a repository, author, date, SHA, or workflow. JSON output, full bodies, open or resolved threads, complete pagination, and a specific CI run are supported. Replies go through pr-reply; this skill does not resolve threads, push changes, or merge pull requests.

An installed and authenticated gh is required, or execution is blocked. A branch without a current PR produces NO_CHANGES for PR-bound work, while conflicting options produce BLOCKED.

What it helps with

  • inspect a pull request and its checks
  • collect every open discussion thread
  • diagnose a failed CI run and locate its logs
  • reply to one pull request conversation
  • run a nontrivial GitHub command through gh

Example prompts for the agent

  • Inspect the current branch PR, including its checks and changed files.
  • Collect all open PR threads with complete pagination and return the result as JSON.
  • Analyze the CI failure for run <run-id> and point me to the log path.

How to install gh

1Skills CLI (Vercel)

npx skills add prathamdby/skills --skill gh -a claude-code

2Manually: put the files in place

~/.claude/skills/gh/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/gh/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about gh

How do I install gh in Claude Code?

Run npx skills add prathamdby/skills --skill gh -a claude-code, then check ~/.claude/skills/gh/SKILL.md.

What does the gh skill do?

It reads pull request state, reviews, discussions, checks, and CI data, and can post a reply to one thread. Covered tasks use dedicated scripts.

What does gh need to run?

The gh binary must be installed and authenticated, and PR-bound requests need a current or explicitly selected PR. Conflicting options block the request.

Source text

SKILL.md · MIT
---
name: gh
description: >
  When the user wants a pull request, its discussion, its checks, or another GitHub record read, or a reply posted without a code change.
---

# GitHub I/O

Contracts follow AVGVSTVS96/better-github-skill (unlicensed; reimplemented).
Resolve `<anchor>` as this skill's absolute directory. The existing
`<anchor>/scripts/run` selects bun, nub, tsx, then Node/nvm with strip-types.

## Options

Default: current branch PR, cwd repo, truncated text. A PR number/URL selects
that PR; a named repo emits -R and needs a PR for PR-bound work.
JSON/full bodies selects --json/--full. Open threads includes unresolved
outdated threads; all includes resolved too; both conflict. Complete paging
adds --complete; author/since supply their filters.
CI selects ci-failures; list runs uses --list (default limit 10), workflow
filters it; a run id analyzes that run. Known head SHA adds --sha.
Reply needs exactly one target/body from `references/reply.md`.
Missing needed values or conflicting choices are BLOCKED.

## 1. Resolve and choose

Confirm authenticated gh; missing binary/auth blocks. Record
`repo | PR/run/SHA | surface | argv | terminal`.
No current PR for PR-bound work → NO_CHANGES.
Choose the covering script:

| Surface | Script | JSON contract |
|---|---|---|
| PR state/checks/files/counts | pr-snapshot.ts | `references/snapshot.md` |
| Review bodies/comments/threads | pr-threads.ts | `references/threads.md` |
| CI runs/jobs/log snippets | ci-failures.ts | `references/ci.md` |
| One native thread/conversation reply | pr-reply.ts | `references/reply.md` |

Snapshot is state, threads is feedback; neither replaces the other.
Before parsing JSON, load only that surface's contract.
Done when target and script or uncovered raw operation are fixed.

## 2. Run

Invoke only through `<anchor>/scripts/run <script.ts> ...`, not node directly.
Node version or ERR_UNKNOWN_FILE_EXTENSION is not failure evidence or a
GraphQL license: use run. Its exit 2 is BLOCKED with tried-runtime list.
User or senior saying "GraphQL is fine" does not skip `run`.
Script success reports data/posts; red CI/open threads are not runtime failure.
For all unresolved feedback use --json --open --complete; SHA-pin known heads.
Replies use pr-reply only. No resolving, pushing, or merging by this leaf.

Raw gh is permitted only for uncovered operations, after applying
`references/raw-gh.md`. A covering-script failure never authorizes reimplementing
its GraphQL. Redirect large output, never pipe gh to head.
For HTTP tracing or logs apply `references/logs.md`.
Done when output is complete, a reply URL exists, or a real error is captured.

## 3. Report

Summarize observed output. Set cap markers mean incomplete data, not success
for an exhaustive consumer. Cite log paths instead of pasting logs.
Green CI, zero threads, or posted reply is SUCCESS; a valid red report is
also successful inspection, not repaired CI.
Terminals: SUCCESS, NO_CHANGES (empty/no PR), BLOCKED.

© prathamdby · MIT · prathamdby/skills

Similar skills

All skills for Git and GitHub →

gh by other authors comes first.

From the same set19

The whole set →
prath-mode prath-mode — a skill for Claude Code, Codex, and Cursor that routes user goals to specific playbooks or delivery chains. It triggers when targeting a single task or end-to-end execution path. CX Codex +1 712installs fix-pr fix-pr — a skill for Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, and OpenCode handles pull request review comments and failed checks. It traces each finding, applies supported fixes, and can push changes when enabled. CX Codex +1 621installs orchestrate The orchestrate skill for Claude Code splits work among subagents; the parent agent sets scope, verifies results, and integrates them without editing product files or delegating its duties. CX Codex +1 618installs recon The recon Claude Code skill reads, creates, or refreshes a saved repository map. It keeps that map outside the project and backs its claims with evidence. CX Codex +1 603installs cursor-agent A skill for Claude Code, Codex, and Cursor runs the local cursor-agent CLI and checks its outcome. It handles tasks involving CLI modes, sessions, and worktrees. CX Codex +1 347installs codex Runs the local Codex CLI when the agent asks. Triggers when a task requires calling Codex CLI with specific flags and no manual approvals. CX Codex +1 331installs claude-code This Claude Code skill runs the local CLI with selected options and checks the result. It also handles requests to change installation, sign-in, or connected tools. CX Codex +1 329installs upfront-design upfront-design — a skill for Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI and OpenCode that runs structured design sessions before any code is written. It pauses after each of five phases until the user approves the output. CX Codex +1 321installs

Project sponsors

Companies that keep this analytics open