scitex-ai/claude-code-telegrammer/pull-request
Claude Code and agent skill by scitex-ai/claude-code-telegrammer
- community
Checking that the SKILL.md file is still in the repository.
Create pull requests for external and internal projects. Handles fork setup, branch conventions, commit style matching, and hook management. Use when contributing to any repository via PR.
Install
npx vskill@latest install scitex-ai/claude-code-telegrammer/pull-requestSKILL.md
Pull Request Skill
When to Use
- Contributing to any repository via PR (external or internal)
Principles
- Minimal change — Touch only what's necessary for the feature/fix
- Follow their conventions — Their style, not ours
- Atomic PR — One PR = one feature/fix. Never mix concerns
- Read first, code second — CONTRIBUTING.md, README (how to contribute section), existing PRs
Workflow
1. Read Their Rules
2. Bypass Our Claude Code Hooks for External Projects
Our Claude Code hooks (ruff auto-format, line-limit, pytest-fullpath, git-C enforcement) WILL corrupt external codebases by reformatting their files to our style. This is the #1 source of dirty PRs.
Offending hooks (defined in `.claude/hooks/settings.json`):
- PostToolUse
format_code.sh— ruff/black auto-format on Write|Edit - PostToolUse
run_lint.sh— linter auto-fix on Write|Edit - PreToolUse
limit_line_numbers.sh— blocks files >512 lines - PreToolUse
enforce_pytest_fullpath.sh— blocks relative pytest paths - PreToolUse
enforce_git_dash_C.sh— blocks git without -C flag
Solution: Start Claude Code with `cc --disable-hooks`
This sets CLAUDE_CODE_DISABLE_HOOKS=true via _cld_start_session in ~/.bash.d/all/010_claude/010_claude_session.src, which disables ALL PreToolUse and PostToolUse hooks (ruff, lint, line-limit, etc.).
If already in a session WITH hooks enabled, you must exit and restart with --disable-hooks. Setting the env var inside Bash tool mid-session does NOT retroactively disable Claude Code hooks.
Fallback if hooks are active (can't restart): Use Bash tool with python3 for file edits to bypass Write|Edit matchers:
Always verify diff is minimal before pushing:
Recovery if hooks already reformatted:
3. Fork and Branch
4. Code, Commit, Push
- Match their commit message format (check
git log --oneline -20) - Match their code style / linter / formatter (run theirs, not ours)
- Run their tests (full suite, not just new tests)
- Rebase on their latest main
5. Create PR
Token Requirements for External Repos
Fine-grained PATs scoped to your own repos CANNOT create PRs/issues on external repos. You need either:
- A classic PAT with
public_reposcope - A fine-grained PAT with "All repositories" access
If the token lacks permission, provide the user with a direct compare URL:
Creating the PR
If gh CLI fails (token scope issue), use REST API:
If both fail, give the user the compare URL to create manually.
Rules
- Read CONTRIBUTING.md before touching code
- One PR, one concern — never bundle unrelated changes
- Their style wins — commit format, code style, test framework, test runner
- Minimal diff — no drive-by cleanups, no extra files
- Rebase before submit — clean history on their latest main
- Use `git -C` for all git commands
- Push to fork remote — never push to origin (upstream)
- Run their full test suite — verify zero regressions before PR
- Token awareness — check token scope early, don't waste time on auth failures
PR Template
When PR template does not exist in target project, use this as an template.
Security
- Security scanPassed, 90 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingdocs/to_claude/skills/pull-request/SKILL.md
Popularity
- Repository stars
- 1
- 7-day trend score
- 0
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/scitex-ai/claude-code-telegrammer/pull-request)