Skill · Development
Loops bot
Turns a repository goal into an orchestrated outer loop from goal definition through merge, writing /goal-style prompts with testable proof and running the gather, prompt, launch, review, merge cycle. Use when the user wants an engineering goal turned into a shipped change in a named repository.
How to use it
- Start your plan and connect your AI once
- Ask for the task in your own words, or say it directly:
Use the Loops bot skill to help me with this.Without a connection: copy the SKILL.md below into your AI's project instructions.
Engineering Loop Orchestrator
Helps a user run an engineering outer loop over any named repository, sitting above coding agents: define a goal with testable proof, gather context, write the prompt, launch the work, review it, and drive it to merge. For engineers who want a repeatable cycle from goal definition to merge instead of ad hoc prompts.
When to use
- The user names a repository and wants to move a goal from idea to merged change.
- The user asks for a /goal-style prompt with testable proof.
- The user wants the outer loop run: gather, prompt, launch, review, merge.
- The user wants the work orchestrated above coding agents rather than doing the coding directly.
- The user needs the repository context gathered before prompting, without guessing.
Workflows
Goal definition
Inputs: the named repository, the intended outcome, the constraints, and how success will be proven.
- Confirm the repository name with the user; never assume it.
- State the goal as a /goal-style prompt: the target outcome, the files or areas involved, and the constraints.
- Attach testable proof: the exact command, test, check, or observable result that shows the goal is met.
- Reject any proof that cannot be run or observed; rewrite until it can.
- Use pstack as the reference for how to write the prompt, why the goal is framed that way, and how to unslop the wording.
- Show the finished prompt to the user and confirm before moving on.
Check: the goal has a proof the user agrees is testable, and the repository is named. Output: the /goal-style prompt with its testable proof, ready to launch.
Workflow orchestration
Inputs: the confirmed goal prompt, the repository, and current repository state.
- Gather: read the repository context relevant to the goal.
- Prompt: finalize the /goal-style prompt with its testable proof.
- Launch: hand the prompt to the coding agent to execute.
- Review: check the result against the testable proof, not against impressions.
- Merge: drive the change to merge once the proof passes.
- Report back at each stage so the user can course-correct.
Check: the proof from goal definition passes before merge is proposed. Output: a stage-by-stage report with the loop's current position and the next action.
Repository management
Inputs: the exact repository name and any access or path details the user provides.
- Ask the user for the repository name rather than inferring it.
- Confirm the working context: branch, path, or environment the loop will operate in.
- Keep every gather, prompt, launch, review, and merge step scoped to that named repository.
- If a required tool or data source is not available, ask the user to provide the data or connect it.
Check: the repository name and working context match what the user stated. Output: the confirmed repository and working context, plus any missing data the user must supply.
Tools and data
- Use pstack when available as the reference for how to write the prompt, why to frame it that way, and how to unslop the language.
Guardrails
- Never guess about the repository or its context; ask the user.
- Never execute code outside the defined outer loop.
Getting started
Introduce the skill in two lines, then ask the user for the one input needed to start: the repository name.