omer-faruk-g

ClaudeG — Development skill for Claude Code

Development community

Claude Code skill: guard rails against weakened tests, unverified 'fixed' claims, placeholders, destructive commands, Windows shell traps and leaked secrets.

How to install ClaudeG

This entry records only its repository, not the path inside it, so there is no exact command to give. Open omer-faruk-g/ClaudeG and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What ClaudeG does

Claude Code skill: guard rails against weakened tests, unverified 'fixed' claims, placeholders, destructive commands, Windows shell traps and leaked secrets.

Alternatives in Development

Careful — Activate destructive command warnings for the session 2.8k ★
Claudecode — Open Source ClaudeCode Leaked 1k ★
Claude On Rails — A development framework for Ruby on Rails developers using Claude Code, inspired by SuperClaude 779 ★

README

ClaudeG

ClaudeG is a [Claude Code](https://code.claude.com) skill that guards against six mistakes coding agents keep making, whatever the model:

Weakening tests to make them pass: deleting or skipping a test, loosening an assertion, special-casing test inputs.
Saying "fixed" without checking: claiming a change works when nothing was run after the edit.
Leaving placeholders: TODO, stubs, NotImplementedError, "rest of the code unchanged", presented as finished work.
Destructive commands: rm -rf, git reset --hard, git push --force, DROP TABLE, run without looking at the target.
Windows shell traps: bash syntax in Windows PowerShell 5.1, /c/... paths and $HOME inside python -c code, CRLF and encoding surprises.
Security slips: API keys written into code, unsafe input handling.

ClaudeG gives Claude rules for all six, plus two small hooks that ask, block or remind where the rules alone slip.

What it does

Part What happens

Rules (SKILL.md) Fix the code, not the test; show a reason and ask before changing a test. Claim success only with a command run this turn and its result line, or say "not verified". No placeholders; list unfinished work. Look before deleting, prefer reversible steps, ask before destructive commands. Windows PowerShell 5.1 / Git Bash / Windows Python / CRLF / encoding rules. No secrets in code or git; safe input handling; no switching off TLS or auth to make something work. A hook decision is final: never reach the same result another way.

Command guard (PreToolUse, Bash, PowerShell) Asks you before rm -r / Remove-Item -Recurse outside temp and build folders, git reset --hard, git clean -f, git push --force / --delete, git checkout --, git restore, git branch -D, git stash drop/clear, history rewrites, find -delete, disk tools, and DROP / TRUNCATE / DELETE without WHERE in database clients. Quoted text, commit message

Part	What happens
Rules (`SKILL.md`)	Fix the code, not the test; show a reason and ask before changing a test. Claim success only with a command run this turn and its result line, or say "not verified". No placeholders; list unfinished work. Look before deleting, prefer reversible steps, ask before destructive commands. Windows PowerShell 5.1 / Git Bash / Windows Python / CRLF / encoding rules. No secrets in code or git; safe input handling; no switching off TLS or auth to make something work. A hook decision is final: never reach the same result another way.
Command guard (`PreToolUse`, `Bash`, `PowerShell`)	Asks you before `rm -r` / `Remove-Item -Recurse` outside temp and build folders, `git reset --hard`, `git clean -f`, `git push --force` / `--delete`, `git checkout --`, `git restore`, `git branch -D`, `git stash drop/clear`, history rewrites, `find -delete`, disk tools, and `DROP` / `TRUNCATE` / `DELETE` without `WHERE` in database clients. Quoted text, commit message