agentleFS
aaroncohnaaroncohn/openspec-conductor-powerups/skills/browser-handoff/SKILL.md

browser-handoff

Skill1 starsChanged 10 days agoMITView on GitHub

Write a self-contained handoff block for a browser task that the user's local Claude Code session (connected to their Chrome with `claude --chrome`) will run. Use when an agent in a Conductor cloud workspace needs a setup or admin task done in the user's real browser (a dashboard setting, a secret pasted into a web UI, a check behind their logins), or when the user asks for a handoff block.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never put a secret in the block.

    line 18
  2. Distinctive:

    Don't upgrade a plan, start a trial, or add a payment method unless the project's rules or the user say so.

    line 19
  3. Distinctive:

    If the task logs in to the project's own app, say how, following the project's rules for logging in.

    line 20

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

browser-handoff

A cloud workspace can't reach the user's browser or its logins. A local Conductor workspace running claude --chrome can. Write the task so that agent can do it with nothing but this block and the repo, then hand the block to the user to paste into that session.

Before you write

Read the project's rules in AGENTS.md / CLAUDE.md (and any docs they point to) for accounts, secrets, plans and billing, and logging in. Carry the ones that apply to this task into the block. The local agent has the repo but not this conversation, so it can read those docs too: link to them instead of copying long sections.

Rules

  • Spell out every name, ID, URL, value and setting the task needs. Don't write "the project" or "the usual account".
  • Name the exact account to use, and have the agent confirm it's signed in as that account before doing anything.
  • Never put a secret in the block. If the task needs one, the block says the user pastes it into the service's web UI themselves.
  • Don't upgrade a plan, start a trial, or add a payment method unless the project's rules or the user say so. Otherwise the agent stops and asks.
  • If the task logs in to the project's own app, say how, following the project's rules for logging in.
  • One task per block. If it touches more than one service or would take more than about 15 minutes, split it.
  • Fill in every section. Write "none" when a section doesn't apply; don't drop it.

Format

Output exactly one fenced block, then one line telling the user where to paste it.

## browser task: <short name>

**Goal:** <one or two sentences: what should be true when this is done, and why>

**Exact account:** <service> as <account>. Before doing anything, confirm the signed-in
account matches. If it doesn't, stop and report which account is signed in.

**Project rules that apply:** <the rules from AGENTS.md/CLAUDE.md that matter here, one line
each, with the doc they come from; or "none">

**Steps:** <numbered, specific: URLs, menu names, field names, exact values>

**Done when:** <checkable criteria: what to see on which page, which value is saved>

**Don't touch:**
- Don't upgrade a plan, start a trial, or add a payment method (unless the project rules above allow it). Stop and ask instead.
- <anything else out of scope: other projects, environments, settings, existing records>
- Confirm with the user before anything hard to reverse or public.

**Report back:** <exactly what to return: IDs, URLs, values that aren't secret, screenshots
saved to `.context/`, or "done" plus anything unexpected>

Paste this into a local Conductor workspace of this repo running claude --chrome.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.