How do I install the Bangle Followup Operator skill?

Run npx skillmds@latest add bangle-io/bangle-followup-operator in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Bangle Followup Operator safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Bangle Followup Operator?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Bangle Followup Operator free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Bangle Followup Operator?

bangle-io (@bangle-io) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Bangle Followup Operator by bangle-io

Name: Bangle Followup Operator
Author: bangle-io

Bangle Followup Operator

Operate recurring Bangle.io agent follow-ups after a thread has already started. Use when the user asks to continue, check status, decide what is next, push, create or update a PR, poll CI, rebase, merge, release, run Playwright/manual UX checks, spawn or aggregate code-review subagents, run thermonuclear review, address review findings, make a plan, add a project task, or asks "we good?", "pushed?", "PR?", "did you test?", "what should I test?", or similar follow-up prompts in the Bangle repo.

bangle-io 7712d4f 2 files · 4.1 KB Updated Sep 17, 2026

File contents

Bangle Follow-up Operator

Overview

Use this skill to convert short Bangle follow-ups into the next concrete action. Keep repository-wide engineering policy in AGENTS.md/CLAUDE.md; keep plan-specific policy in plans/AGENTS.md.

Reconstruct State

Read the latest prompt and recent conversation turns using the thread/history tools available in the current agent environment.
Check repository state from the root: branch, dirty files, remotes, active PR, recent commits, and any generated worktree context.
Infer terse prompts like continue, what next, we good, pushed?, or PR? from the immediately previous work before asking for clarification.

Route the Follow-up

Continue or status

Inspect unfinished commands, partial edits, unpushed commits, failed checks, and PR/CI state.
Report exact current state, then continue the next unfinished step.

Review or subagent request

For thermonuclear review, use the thermonuclear-code-quality-review skill.
For normal review or subagent review, pass raw scope: repo path, branch/PR, diff target, constraints, and whether the agent may edit.
Aggregate findings into must fix, worth fixing, and defer/track.

Address findings or "do it"

Re-read the cited code and tests before editing.
Implement only fixes that match the user's scope and improve the branch.
Prefer changes that reduce total uncertainty and complexity. Type and lint cleanup has strong ROI when it clarifies a real boundary, improves useful caller inference, or makes an invalid state unrepresentable. Reconsider it when satisfying the rule adds conditional type machinery, assertions, or downstream churn without a concrete new guarantee. Type-level wildcards and runtime boundary leaks can look similar in a scan but have different risk; judge the resulting API and implementation, not the warning count alone.
After edits, run focused checks first, then the required AGENTS.md gates for the changed surface.

UI, UX, or Playwright follow-up

Use playwright-cli, browser, or Chrome tools when the user asks for manual UX validation, screenshots, authenticated Chrome, or visual comparison.
If the user points at production, t3code, prosekit, or another local reference, inspect that reference before claiming parity.

PR, git, CI, and deploy follow-up

Before push/PR/merge/release, verify branch, remote, dirty state, and target PR.
After push, provide the PR URL or branch and poll relevant GitHub checks when requested.
For Cloudflare preview or GitHub Action comments, verify whether displayed commit hashes are PR head SHAs, merge SHAs, or deployment branch SHAs before patching.
For releases, use the applicable Bangle release skill and AGENTS.md release gates.

Planning or project task

For durable plans under plans/, read plans/AGENTS.md first.
For adding, updating, triaging, scheduling, or reconciling Bangle project items, use bangle-project-operator.
When a plan is fully done, move it to plans/archived/, set status: completed, set archived: true, set archived_on: YYYY-MM-DD in the YAML frontmatter, update updated, and include the DONE note required by plans/AGENTS.md.

Close Out

State what changed or what was found.
Include branch/PR/push/CI status when relevant.
Include checks run and any checks intentionally not run.

---name: bangle-followup-operatordescription: Operate recurring Bangle.io agent follow-ups after a thread has already started. Use when the user asks to continue, check status, decide what is next, push, create or update a PR, poll CI, rebase, merge, release, run Playwright/manual UX checks, spawn or aggregate code-review subagents, run thermonuclear review, address review findings, make a plan, add a project task, or asks "we good?", "pushed?", "PR?", "did you test?", "what should I test?", or similar follow-up prompts in the Bangle repo.---# Bangle Follow-up Operator## OverviewUse this skill to convert short Bangle follow-ups into the next concrete action. Keep repository-wide engineering policy in `AGENTS.md`/`CLAUDE.md`; keep plan-specific policy in `plans/AGENTS.md`.## Reconstruct State1. Read the latest prompt and recent conversation turns using the thread/history tools available in the current agent environment.2. Check repository state from the root: branch, dirty files, remotes, active PR, recent commits, and any generated worktree context.3. Infer terse prompts like `continue`, `what next`, `we good`, `pushed?`, or `PR?` from the immediately previous work before asking for clarification.## Route the Follow-up### Continue or status- Inspect unfinished commands, partial edits, unpushed commits, failed checks, and PR/CI state.- Report exact current state, then continue the next unfinished step.### Review or subagent request- For `thermonuclear review`, use the `thermonuclear-code-quality-review` skill.- For normal review or subagent review, pass raw scope: repo path, branch/PR, diff target, constraints, and whether the agent may edit.- Aggregate findings into `must fix`, `worth fixing`, and `defer/track`.### Address findings or "do it"- Re-read the cited code and tests before editing.- Implement only fixes that match the user's scope and improve the branch.- Prefer changes that reduce total uncertainty and complexity. Type and lint cleanup has strong ROI when it clarifies a real boundary, improves useful caller inference, or makes an invalid state unrepresentable. Reconsider it when satisfying the rule adds conditional type machinery, assertions, or downstream churn without a concrete new guarantee. Type-level wildcards and runtime boundary leaks can look similar in a scan but have different risk; judge the resulting API and implementation, not the warning count alone.- After edits, run focused checks first, then the required `AGENTS.md` gates for the changed surface.### UI, UX, or Playwright follow-up- Use `playwright-cli`, browser, or Chrome tools when the user asks for manual UX validation, screenshots, authenticated Chrome, or visual comparison.- If the user points at production, t3code, prosekit, or another local reference, inspect that reference before claiming parity.### PR, git, CI, and deploy follow-up- Before push/PR/merge/release, verify branch, remote, dirty state, and target PR.- After push, provide the PR URL or branch and poll relevant GitHub checks when requested.- For Cloudflare preview or GitHub Action comments, verify whether displayed commit hashes are PR head SHAs, merge SHAs, or deployment branch SHAs before patching.- For releases, use the applicable Bangle release skill and `AGENTS.md` release gates.### Planning or project task- For durable plans under `plans/`, read `plans/AGENTS.md` first.- For adding, updating, triaging, scheduling, or reconciling Bangle project items, use `bangle-project-operator`.- When a plan is fully done, move it to `plans/archived/`, set `status: completed`, set `archived: true`, set `archived_on: YYYY-MM-DD` in the YAML frontmatter, update `updated`, and include the DONE note required by `plans/AGENTS.md`.## Close Out- State what changed or what was found.- Include branch/PR/push/CI status when relevant.- Include checks run and any checks intentionally not run.

Frequently asked questions