extrovert-sdk
Skill0 starsChanged 2 months agoMITView on GitHub
Build or troubleshoot a TypeScript integration against the Extrovert REST API using the current SDK source and OpenAPI contract. Use for delegated administrative workflows, client construction, org/project/inbox addressing, thread workflows, pagination, error handling, commerce requests, reviewed sends, OTP waits, webhooks, streams, signature verification, or offline fixture tests.
- Installs packages
In this repository10 other files it gives its agents.
- agent-email-architectureSkill
- configure-himalayaSkill
- extrovert-adminSkill
- extrovert-connectSkill
- extrovert-manage-inboxesSkill
- extrovert-read-inboxSkill
- extrovert-send-emailSkill
- extrovert-supportSkill
- extrovert-writing-rulesSkill
- wait-for-otpSkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Read the live guide for current product behavior and use the host's current tool schemas.
- Very specific: line 55
Keep the SDK, current OpenAPI, and backend behavior in lockstep.
≈ “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.
Files
Extrovert TypeScript SDK
Check current guidance
On first Extrovert use in this session, after one hour (or a shorter returned freshness interval),
and after an unknown-tool or schema error, inspect the current tool catalog. Call agent_context
only if present; otherwise fetch
https://mcp.extrovert.dev/.well-known/agent-contract.json, then https://docs.extrovert.dev/llms.txt.
An installed or pinned CLI can run agent status --json if supported. Only when normal installation
policy permits an unpinned CLI, use
npx --yes --prefer-online @extrovert.dev/mcp agent status --json.
Read the live guide for current product behavior and use the host's current tool schemas. If a
schema remains stale, refresh the catalog or reconnect before continuing; inspect state before
retrying an uncertain mutation.
Compare this skill's metadata.version with its version in live context. A difference signals a
refresh to consider, not incompatibility, permission to downgrade, or authorization to install.
Preserve explicit pins, local edits, the installation manager, and scope; refresh only the installed
Extrovert skills when permitted. Updating files requires two separate reloads: reload and re-read each changed SKILL.md, then
restart changed local MCP processes or refresh hosted discovery. If a new session is needed, report
that pending step; use live guidance for this task. If freshness is unavailable, report
that condition without treating it as disabled signup or permission to guess new behavior. See
updates for targeted refresh instructions.
The prerelease package is published under the explicit next dist-tag. Use
npm install @extrovert.dev/sdk, or pin @extrovert.dev/sdk@0.1.0-pre.13 for reproducible dogfood
tests. The public REST API and served OpenAPI remain the underlying contract. Do not imply that a
hosted MCP route substitutes for the SDK; use the SDK for direct TypeScript application integration.
Choose the smallest mode
Single-inbox reads expose human_email_review: the default-off account-wide option
to skip review for exactly one To recipient at the verified human email, no Cc/Bcc
or aliases. All writing rules, intent and sending limits still apply; other recipients
keep their policy. The signup practice always requires review. Ordinary credentials
can read this but cannot change it. Explicit Full account control can use the
administrative review-policy PATCH with skip_review_for_human; omitted fields are
unchanged and null is invalid. Use the returned settings_url for relevant discovery,
not repeated prompts. See the live agent contract for details.
- Basic client, authentication, inbox handles, thread reads/search, project addressing, pagination, metadata, commerce requests, or errors: read references/core-client.md.
- Customer administration, credential handoff, and revocation: read references/administration.md.
- Sending, replying,- Sending, replying, forwarding, durable review events, revisions, or rules: read references/review-loop.md.
- Webhooks, signature verification, SSE, or offline fixtures: read references/events-and-testing.md.
Use whoami() as the authority source. Prefer the canonical projects.inboxes.* chain when project breadth matters; inbox ids and cursors are opaque. A body project_id asserts the active project. Legacy keys keep their fixed project; a connection can select a project only within its consented reach. Use whoami.connection when present instead of applying legacy ownership assumptions.
Keep the SDK, current OpenAPI, and backend behavior in lockstep. Parse typed problem responses rather than branching on message strings. Pass stable idempotency_key values for retryable mutations, and never report a queued review as delivered.
Internal email review exceptions
Read internal_email_review in inbox detail alongside the existing human-recipient exception. Project and organization exceptions start off. Every final To/Cc/Bcc recipient must resolve to an inbox in the sender's enabled project or exact organization. Shared domains, aliases, shared humans, child organizations, and mixed external recipients do not qualify. Organization enablement includes current and future projects; project-off does not override it. These settings grant no inbox access.
Ordinary agents and project managers cannot enable these settings. A current owner/admin or explicitly delegated Full account control can change them using the existing administrative review-policy operation. Explain a returned settings URL when relevant, but do not repeatedly solicit enablement, widen your authority, or split recipients to evade review. Apply writing rules and intent, then submit normally and follow the returned sent/queued result. Sending limits and protected onboarding review remain in force. See https://docs.extrovert.dev/review-loop/agent-contract/#email-between-agents-without-review.
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
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.

