# Contributing Thanks for helping make Grok Build easier to use. ## What We Want High-signal contributions: - official xAI links and doc updates; - working `.grok/skills`; - safe hook patterns; - AGENTS.md templates for real project types; - stack packs under `templates//`; - concise examples of prompts that produced useful diffs; - community projects with public code or demos; - comparisons based on actual usage, not vibes. ## Contribution Rules - Label sources as `official`, `community`, `experimental`, or `commercial`. - Do not submit private Discord screenshots or paywalled content unless you have permission. - Do not submit secrets, API keys, logs with tokens, or private repo data. - Avoid generic "AI will change everything" prose. Show concrete workflow value. - Keep Markdown tables readable on GitHub. ## Adding a Community Project Add it to `README.md` with: - project name; - URL; - one-line description; - status: official, community, experimental, or commercial; - why Grok Build users should care. ## Adding a Skill Create: ```text .grok/skills//SKILL.md ``` Use this frontmatter: ```yaml --- name: skill-name description: Use when... version: 1.0.0 author: your-name-or-org --- ``` Keep the skill narrow. Strong Grok Build skills include: - Plan Mode guidance; - subagent decomposition for larger tasks; - Arena-style comparison when multiple approaches are useful; - web/X search guidance when current information matters; - human approval checkpoints before risky edits; - lint/test verification guidance; - git hygiene. If it needs examples, add: ```text .grok/skills//examples/ ``` ## Skill Quality Checklist - [ ] The description says exactly when the skill should trigger. - [ ] The skill is useful outside your own repo. - [ ] It does not hide destructive commands. - [ ] It separates confirmed facts from community observations. - [ ] It tells the agent how to verify work. - [ ] It explains when to ask for human approval. - [ ] It keeps `SKILL.md` concise and moves bulky material into `references/` or `examples/`. ## Pull Request Checklist - [ ] I verified links I added or changed. - [ ] I clearly marked unofficial or experimental projects. - [ ] I did not add secrets or private data. - [ ] I kept claims specific and dated where needed. - [ ] I updated docs/examples if behavior changed. - [ ] New skills include `name`, `description`, `version`, and `author`. - [ ] New skills mention safety and verification when they run commands. ## Code of Conduct Be useful, direct, and generous. Debate claims and evidence, not people.