agentleFS
matthewdonsemail-labmatthewdonsemail-lab/log/.claude/skills/log-inbox-sign-in/SKILL.md

log-inbox-sign-in

Skill63 starsChanged 3 days agoMITView on GitHub

Signing an account in: checking it, then the live browser. A log play (docs/useCases/inbox/signIn.md): An account's session has run out, or it was never signed in. The agent checks it through the platform's own answer, and when it needs a person, it opens the account's own browser for the owner to sign in. On the owner's machine that browser is the window on their screen. From a remote runtime it is streamed into the chat as a live view, the same one the console's Connections page opens. The agent never types a password; the owner does.

The file

Files
SKILL.md

Signing an account in: checking it, then the live browser

This is the log play inbox/signIn. Before anything else, call the log MCP tool getGuide({ useCase: "inbox/signIn" }): it returns this page and makes it the session's play, so every log tool's answer says which step you are on and a send with no play loaded is refused (rules 157, 158). Then follow the steps in order.

Draft every message from that person's own thread and matt's real replies (voice with no text, search), never from memory, and never send two that read alike (sameAsSent).

An account's session has run out, or it was never signed in. The agent checks it through the platform's own answer, and when it needs a person, it opens the account's own browser for the owner to sign in. On the owner's machine that browser is the window on their screen. From a remote runtime it is streamed into the chat as a live view, the same one the console's Connections page opens. The agent never types a password; the owner does.

How the owner asks for it: "sign in to my account", "is my x account logged in", "check my accounts", "log me back into instagram", "my session expired", "check the status of my accounts", "open the browser so I can sign in"

Run it from a fresh context

Say: "go on docs/useCases/inbox/signIn.md for <account>". The agent reads this page and runs the steps.

The steps

#StepToolDone when
1Check every account (or one) through its platform's own answer, never a screen. A 401 or no saved session is "needs signing in", with its next step. An account held stopped is never checked again (rule 94)sessionsCheck { platform?, account? } (the CLI's log sessions check, the console's Check)each account signed in, needs signing in, stopped or unknown
2Sign it in with sessionsSignIn: it checks the account first, then opens the account's own browser on the platform's login screen (X, with no browser yet, takes the masked cookie prompt in a terminal). browserView shows that open browser: on the owner's machine the window on their screen; from a remote runtime an MCP Apps host shows it inline in the chatsessionsSignIn { account, platform? }, then browserView { account }the answer is waitingForOwner, and says where the browser is
3The owner signs in themself, in the window or in the chat. Their clicks and typing reach the page only while it still asks for sign-inthe ownerthe page has moved past sign-in
4Finish it: sessionsSignIn with done closes the browser, saves its session only when the platform's own sign-in cookies are in it, and checks again. Only call it signed in when the platform says so (rule 25)sessionsSignIn { account, done: true }signed in

What to know

  • One sign-in, everywhere. sessionsSignIn, log sessions signIn and the console's Sign in and Browser check run the same machine (src/platformKit/signIn), and the console's Connections page and browserView use the same live browser, through src/browser/viewer. Every check is recorded against the account, so all of them show the same state.
  • A signed-out browser saves nothing. Closing the browser keeps the stored session unless the platform's own sign-in cookies are there; the answer says what was saved.
  • Where it shows. Inline sign-in needs an MCP Apps host (Claude, ChatGPT, Cursor). On the owner's own machine there is no stream: the window is on their screen. Any other host gets the text answer, which says where to sign in.
  • A checkpoint is a stop. A captcha, a code the platform sends, or a security check is the owner's to answer. The agent waits and checks again, and never retries around it (rule 94).
  • Passwords stay with the owner. The agent opens the browser and checks the result; it never types a password or a code.

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.