agentleFS
extrovert-dot-devextrovert-dot-dev/extrovert-skills/extrovert-connect/SKILL.md

extrovert-connect

Skill0 starsChanged 2 months agoMITView on GitHub

Connect an existing Extrovert account, resume signup, or diagnose identity and access. Use for OAuth, enrollment, host setup, and connection recovery; broader administration needs explicit consent.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do not create another account to repair a connection or start a test email without permission.

    line 12
  2. Very specific:

    If this host already has Extrovert tools, use their current schemas.

    line 17
  3. Very specific:

    Prefer the host's remote MCP/OAuth connection when supported: it requires no local Node, compiler, SDK, or source checkout.

    line 22
  4. Distinctive:

    If remote MCP is unavailable, reuse an installed CLI.

    line 26
  5. Very specific:

    Call whoami in the actual connection, then list_inboxes and get_inbox for its intended inbox.

    line 50
  6. Very specific:

    After updating local MCP, reload this host connection (Hermes: /reload-mcp) and verify another MCP call here.

    line 89

≈ “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

Connect to Extrovert

Reuse working access. A console handoff means an existing account; missing local credentials do not mean a new customer. Ask about account intent only when unclear. Do not create another account to repair a connection or start a test email without permission. A pending signup takes precedence over a new browser login.

Choose the shortest working path

  1. If this host already has Extrovert tools, use their current schemas. On first use, after the returned freshness interval (normally one hour), or after a schema error, inspect the catalog and call agent_context if available. Without it, fetch https://mcp.extrovert.dev/.well-known/agent-contract.json and use https://docs.extrovert.dev/llms.txt to select only the needed guide.
  2. Prefer the host's remote MCP/OAuth connection when supported: it requires no local Node, compiler, SDK, or source checkout. For ordinary existing-account email use https://mcp.extrovert.dev/assistant/mcp. Use an already installed plugin's endpoint unchanged. Installation of skills alone does not connect tools.
  3. If remote MCP is unavailable, reuse an installed CLI. Its published package needs Node >=20 but no TypeScript compiler, pnpm, Git, or Docker. Read installation only to install, update, or repair the host. Do not install a runtime merely because a different path mentioned it.
  4. An agent with secure HTTPS execution but no MCP/Node can use the HTTPS recipes. If the host has neither authenticated tools nor executable HTTPS, hand off the specific human step. Do not invent a transport or confuse the human's laptop with the agent runtime.

Connect the intended identity

For existing accounts, authorize through this host's native OAuth action. Local CLI/stdio auth login is an alternative, not a second required login and not proof of a hosted session. Give the sign-in URL/action to the person in this chat; never infer their Extrovert email from a local or model-provider account.

Recommend Dedicated agent, Selected inboxes, and Read and send for ordinary email. Identity, resource reach, and actions are separate. Selected inboxes excludes future inboxes. Creation needs appropriate project/organization reach and mailbox:create, not Full account control. Never widen access to fix an empty list. For credential enrollment, worker delegation, expiry, or denied access, read access and diagnostics. Full account control requires an explicit administration request; created credentials can outlive that connection.

Call whoami in the actual connection, then list_inboxes and get_inbox for its intended inbox. Lead with account/project names and capabilities; retain IDs, scopes, reach and expiry privately. A saved configuration, successful browser callback, or CLI health check does not prove hosted MCP access. After repeated 401s at the same stage, stop approvals and report its sanitized error/request ID.

In Codex, codex mcp add extrovert --url https://mcp.extrovert.dev/assistant/mcp starts authorization itself; do not begin a second concurrent login. If that flow needs recovery, run codex mcp login extrovert --scopes extrovert:connect, then verify with native whoami and an authorized inbox read. An "Unknown" CLI auth label is inconclusive when those authenticated calls work. A release does not require reconnecting a working hosted connection.

Resume, then do useful work

  • New account or pending signup:verify credential: read signup before acting. Preserve the original account, profile and expiry. Pass an optional gift code with human_email; a pending gift requires the human to claim that same workspace without a card. Wait for actual human proof; never invent a code or approve it.
  • Existing authorized send or "any feedback?": use list_reviews with composer: "me" and list_review_events, then load extrovert-send-email and extrovert-writing-rules. Resume the same draft. Inspection-only means reads only.
  • Missing inboxes, domain readiness, capacity, or event delivery: read inboxes and events only for that problem.

Report the proven milestone and next action: connected identity, draft awaiting review, or confirmed sent message. A claimed inbox still needs credential exchange; a queued draft is unsent; sent is not proof of receipt. A stopped host cannot wake itself by polling. Preserve secrets, existing pins, installation scope and retry IDs.

Support and runtime checks

For support lists and case-ID reads/replies, use the authenticated connection directly; no preliminary whoami or project lookup is required. See extrovert-support. For update verification, inspect executing_runtime in an MCP agent_context or whoami call from this conversation. If executing_runtime is absent, report that this connection's running version is unknown. Older tools show an "Extrovert VERSION" heading from the hosted release; neither that heading, release_version, tool count, nor installed skill versions proves the executing MCP package version. Do not claim the update succeeded or failed from those values. A fresh CLI reports only its own invocation. After updating local MCP, reload this host connection (Hermes: /reload-mcp) and verify another MCP call here. Preserve pins, profiles and edited skills.

For a new inbox, lead with email activation and briefly offer same-email console signup as the alternative. Console registration links the existing workspace; it is not a second agent signup. See signup for the concise human handoff and active watch.

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.