agentleFS
Sign inStart free
darellchua2darellchua2/civiltekk-opencode-claude-skills/skills/browser-fallback-skill/SKILL.md

browser-fallback-skill

Skill7 starsChanged 56 days agoApache-2.0View on GitHub

Route browser work across agent toolsets — probe a managed desktop-browser toolset read-only before any mutating call, fall back to independent automation stacks on disconnect, verify-then-handoff for show-the-user intents. Harness values in references/. Triggers: browser.disconnected, no desktop browser connected, browser toolset, probe before tabs.open, desktop browser attach, browser fallback.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    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:

    Before the first call to a managed desktop-browser toolset in any task.

    line 30
  2. Distinctive:

    After any "no desktop browser connected" / client-not-attached failure.

    line 31
  3. Distinctive:

    When deciding which browser stack serves a given intent.

    line 32
  4. Distinctive:

    Not for: e2e test-suite authoring (use the project's test conventions).

    line 34
  5. Distinctive:

    Gate: no browser-automation toolset in the session catalog → inert, stop.

    line 38
  6. Distinctive:

    If unsure whether a call is read-only, treat it as mutating and find a cheaper probe.

    line 56
  7. Distinctive:

    Never build detection on undocumented env or interface signals — they vanish without notice.

    line 57

≈ “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

Skill: browser-fallback-skill

What I do

  • Gate on toolset presence: a managed desktop-browser toolset appearing in the session catalog does NOT mean its backing client is attached.
  • Probe the managed toolset read-only before any mutating call — never let an open/preview call be the first touch.
  • On a disconnect verdict: never retry (retries cannot attach the client); fall back to independent browser automation stacks.
  • Show-the-user intents: verify the page via the fallback stack (load check + screenshot), report what was seen, hand off the URL plus how to attach the client.
  • Automation intents (screenshots, DOM/console inspection, viewport checks): route straight to an independent stack; skip the managed toolset entirely.
  • Re-probe at use time — catalogs and client attachments change mid-session; never latch.

When to use me

  • Before the first call to a managed desktop-browser toolset in any task.
  • After any "no desktop browser connected" / client-not-attached failure.
  • When deciding which browser stack serves a given intent.

Not for: e2e test-suite authoring (use the project's test conventions); MCP server installation; tool permission configuration.

Method (decision rules)

  1. Gate: no browser-automation toolset in the session catalog → inert, stop. Catalog presence is necessary, never sufficient.
  2. Classify intent: show-the-user (the user must see it) vs automation (the agent verifies or inspects).
  3. Automation intent → independent stack directly. Done.
  4. Show-the-user intent → read-only probe of the managed toolset FIRST — cheap, side-effect free, never a mutating call.
  5. Probe reports disconnect → no retry. Load the harness side file (table below) for that harness's probe and fallback pairs; verify the target via the fallback stack; report findings; give the user the URL and the attach step.
  6. Long task → re-probe at each new use; attachment is session-lifetime-mutable.

Harness values (load rule)

ReadWhenUse
references/opencode.mdsession runs in OpenCode, or the managed toolset is browser.*probe method, mutating first-touches to avoid, disconnect signature, fallback stack pairs, doc citations

Another harness with a managed browser toolset and no side file: discover locally — identify one read-only catalog call, the mutating calls, and the installed independent stacks — then contribute references/<harness>.md copying this file's skeleton.

Hard rules

  • One probe per intent resolution; a disconnect verdict stands for the task unless the user says they attached the client.
  • If unsure whether a call is read-only, treat it as mutating and find a cheaper probe.
  • Never build detection on undocumented env or interface signals — they vanish without notice; the probe is the contract.

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.

Check your setup, free