Five coding workflows
Grok Bot templates for GitHub and repository work
GitHub Trending Scout
Surfaces trending GitHub repositories each morning and explains in Chinese why each is worth a look.
Choose it to monitor trending GitHub repositories and get a short explanation of why each one deserves attention.
View templateSEOAgent
Sets up SEOAgent in a website repository, evaluates SEO recommendations, and applies changes with recorded reasons.
Fits a website repository that needs SEO recommendations applied with a record of what changed and why.
View templateloops
Runs an engineering outer loop above coding agents, writing testable prompts and checking their work against a reference stack.
Runs an engineering loop above coding agents, with testable prompts and a review step against the reference stack.
View templateFable 5.1 Oracle
Uses Claude Code to plan repository work and review results, producing handoff packets for a separate coding agent.
Plans repository work and reviews the result before handing a compact packet to a separate coding agent.
View templateAgent Looper
Keeps a local coding agent iterating until your acceptance test passes
Keeps a local coding agent iterating until the acceptance test you provide passes.
View templateChoose the template for the stage of work
A GitHub scout, repository editor, coding orchestrator, and review Bot solve different problems. Start with the bottleneck you have now instead of giving one Bot every engineering responsibility.
- Discovery: watch repositories, releases, or open-source trends.
- Planning: turn an issue or brief into bounded repository work.
- Execution: delegate a coding task with a clear acceptance test.
- Review: inspect the diff, test results, and remaining risks before merge.
Give repository access carefully
Check whether the template expects local files, GitHub access, a command line, or another coding agent. Start in a branch or worktree, keep credentials out of prompts, and review commands before they run against important environments.
Define proof before the coding starts
State the expected behavior, affected files, constraints, tests, and the point where the Bot must stop for review. A precise completion check is more useful than asking it to keep improving the code until it feels done.
Keep the handoff reviewable
Ask for the changed files, test output, unresolved risks, and any decisions the next person or Bot must make. Repository history should show the work clearly enough that another developer can continue without reconstructing the conversation.