AGENT SKILL

curviate-quickstart

The first run: install the Curviate CLI, authenticate this machine with `curviate setup`, confirm with `curviate doctor`, connect a LinkedIn account, and prove the whole path works by reading the same profile live and then from the store. Use when Curviate is being set up for the first time, when `curviate` is not installed or not authenticated, when a command fails with a credential or account error, or when someone asks how to get started. Reads only: it issues no write of any kind.

License 许可
MIT
最近更新
2026-11-11
Stars
★ 3.3w
Forks
3.9k

01 前言 Front Matter

name
curviate-quickstart
license
MIT
metadata
{ "tags": [ "LinkedIn", "CLI", "Agents", "Sales", "Recruiting", "Outreach" ], "author": "Curviate", "version": "0.1.0", "repository": "https://github.com/Curviate/curviate-plugin" }
description
The first run: install the Curviate CLI, authenticate this machine with `curviate setup`, confirm with `curviate doctor`, connect a LinkedIn account, and prove the whole path works by reading the same profile live and then from the store. Use when Curviate is being set up for the first time, when `curviate` is not installed or not authenticated, when a command fails with a credential or account error, or when someone asks how to get started. Reads only: it issues no write of any kind.

03 SKILL.md 正文

Curviate: the first run

Four steps, in this order: install, authenticate, connect an account, prove it. Every command here is a read. Nothing in this skill posts, messages, invites, follows or endorses anyone.

Established against CLI 0.33.0.

1. Install the CLI

npm install -g @curviate/cli
curviate --version    # needs Node 18 or newer

2. Authenticate this machine: curviate setup

setup opens the dashboard, takes the short code it shows you, and saves an API key into a local profile. The key is never printed and never passed on a command line.

The command has two shapes, and which one you get is decided by whether stdout is a terminal.

Agent shape: two legs, because a browser has to happen in between

When stdout is not a terminal, or when --json is passed, setup prints a JSON object and exits 0. It does not prompt and it does not open a browser:

curviate setup --json
# {"authorize_url":"…","next_step":"curviate setup --code -","instructions":"…"}

Open authorize_url in a browser signed in to the workspace, press Authorize, and read back the code it displays. Then hand that code to the second leg on stdin:

printf '%s' "$CODE" | curviate setup --code -
# {"ok":true,"tenant":"…","profile":"default","account_id":"…"}
  • --code - reads from stdin deliberately. Passing the code as a flag value works but warns: a value on the command line is visible to other processes through ps and is saved in shell history.
  • The two legs must run on the same machine, under the same configuration directory. Leg one writes a short-lived resume file beside the config file, readable only by you, holding the session id and the private key that opens the delivered credential. Leg two reads it and deletes it. Split the legs across machines, users, containers or config-directory overrides and the exchange cannot complete.
  • --code - with no setup in progress exits 2 and says so. Run leg one first.
  • The code is short, case-insensitive, and drops anything that is not one of its own characters, so a pasted code with stray spaces or punctuation is still accepted.

Human shape

With stdout on a terminal and no --json, setup prints the link, opens a browser, and prompts for the code (up to three attempts). --no-browser prints the link rather than opening it, which is what you want over SSH or anywhere a browser would open on the wrong machine.

One asymmetry to know before scripting it: stdout on a terminal with stdin piped is refused at exit 2, because there is nowhere to paste a code into. It names CURVIATE_API_KEY and curviate login --api-key - as the alternatives. A fully piped invocation is not that case: it is the agent shape above, and it exits 0.

What each failure means

Exit Cause What to do
0 Authenticated. The key is saved to the named profile. Continue to doctor.
2 Usage: no setup in progress, or a terminal was needed and there was none. Fix the invocation. Never retry unchanged.
3 The code was wrong, expired, or the attempts ran out; or the workspace has no API key to deliver; or the key saved but a verifying call did not succeed. Read the message: these are deliberately four different sentences, and only the first is fixed by starting over. Wrong or expired code: run setup again for a fresh one. No API key: create one in the dashboard first. Saved but unverified: run doctor.
6 Too many attempts against this endpoint. Wait, then try again.
7 The API could not be reached. Check the network, and the base URL if you overrode it.
1 The exchange failed in a way the CLI did not expect, or the delivered credential could not be opened by this process. Worth one retry. If it repeats it is a bug to report, not a state to work around.

A code is short-lived and allows only a handful of attempts. If in doubt, start a fresh setup rather than reusing anything.

3. Confirm: curviate doctor

curviate doctor --json

It reports the CLI version, the config path, the active profile, the base URL, where the credential resolved from, which workspace it authenticated as, whether the API is reachable, whether the credential is accepted, and every connected account with its status.

Three things that surprise people:

  • Run setup first, then doctor, not the other way round. Reachability is only tested through a real call, so with no credential resolved doctor reports the API as unreachable. That is not a network problem; there was nothing to call with.
  • The workspace line is populated only for a credential setup wrote. A key supplied through the environment or login authenticates perfectly well, but doctor has no name to show for it and says so. And where a name is shown it identifies the workspace; it is not promised to be a friendly, human-chosen label.
  • Zero connected accounts is a pass, not a failure. doctor reports 0 connected and exits 0. That is the expected state of a workspace nobody has connected an account to yet.

doctor reports rather than throws, so it prints a report and not an error envelope. Branch on the exit code and read checks[] for which stage failed:

  • 0: every check passed.
  • 2: the request was refused before it was sent, so nothing reached the network. An empty credential, or a malformed base URL (the report's detail reads Invalid base URL: expected an absolute http:// or https:// URL). Fix the invocation or the value; a retry cannot help.
  • 3: no credential resolved, or one resolved and was rejected. Run setup.
  • 7: the API could not be reached, or was reached and answered with a platform fault. Both are worth a retry.

This list is not exhaustive, and the codes beyond it are not doctor's own. The credential check is a real API call, and whatever refusal comes back is passed straight through, so any code the API can produce can surface here. The one a first run meets most often is exit 5 on a workspace with no active seat: the credential is fine and the network is fine, and no amount of re-running setup will change it. Look an unfamiliar code up in the area skill for the surface you are about to use: each ends with a table of the codes its commands produce and the action each implies.

The checks[] entries separate the three stages, and the wording is exact: api reachable reading not checked means the request never left this machine, while could not reach means it left and found nothing. credential valid reading not checked means nothing asked the credential anything; it is not a verdict on the credential, so do not go looking for a bad key on the strength of it.

4. Connect a LinkedIn account

Curviate acts on LinkedIn through an account you connect. Until one is connected, authentication is complete and correct while every account-scoped command is unusable: that is the whole state, and it is not an error to be debugged.

curviate account list --json    # empty on a fresh workspace

If setup reported an account_id, an account is already connected and already set as this profile's default; skip the rest of this step, including the block below, and go to step 5.

Otherwise, find a free seat first:

curviate account seats --json    # seat_id of any entry with "occupied": false

Then connect one with curviate account link --seat-id <id> .... It attaches a LinkedIn account to that seat and usually needs a verification code, which means a human. Non-interactive shells get exit 12 and finish through the checkpoint flow. The command surface, its flags and the checkpoint commands are in curviate-profile: read that before running it, because a half-finished connect leaves a session to poll rather than a clean failure.

Then tell the CLI which account to act as:

curviate account list --json              # the acc_id of the account you just connected
curviate config set-account <acc_id>      # make it this profile's default

Do not skip this. setup stores a default account only when the workspace already had one to hand back, which is exactly what the expected first run does not have. Without a default, every account-scoped command, including the demonstration below, stops at exit 2 asking for --account. Set it once here, or pass --account <acc_id> on each call. config set-account writes to the stored profile, so it needs one: if your credential came from CURVIATE_API_KEY rather than setup, it exits 2 saying there is no config file, and --account <acc_id> per call (or CURVIATE_ACCOUNT in the environment) is the path for you. Account ids change when an account is reconnected, so read one live rather than hard-coding it.

5. Prove the whole path: the same read, twice

This is the demonstration, and it is worth running because it exercises authentication, the connected account, projection and the store in one pass. Both calls are reads of your own profile.

curviate profile me --mode live --json          # "source":"live"
curviate profile me --mode cache_only --json    # "source":"store", with "observed_at"

The first forces a live call to LinkedIn. The second refuses to go to LinkedIn at all and answers from the copy the first one stored, so it comes back with source: store and an observed_at timestamp saying when that copy was taken.

Read source rather than assuming it. That field, not the freshness you asked for, is what actually happened. The full set of modes, and what a cache_only read does when nothing is stored, are in curviate-profile.

Full command surface

<!-- generated: command surface, CLI 0.33.0 -->

Read from the CLI's own --help at version 0.33.0. Descriptions, traps and confidence tags elsewhere in this skill are hand-written and carry the version they were established against.

Every command below that takes flags at all also accepts --base-url, --json, --profile.

Command Arguments Flags
curviate setup (none) --no-browser, --code
curviate doctor (none) --api-key, --timeout
curviate login (none) --api-key, --account

<!-- /generated -->

Where to go next

Back to curviate for the dispatch table. The common first tasks: look someone up (curviate-profile), find a set of people (curviate-search), read your inbox (curviate-inbox). Each area skill ends with the exit codes its commands can produce and what each one implies; branch on those, never on the message text.

04 同仓库更多 Skills

davila7
davila7/claude-code-templates

wealthville-pool-scoring

Score DeFi liquidity pools before providing liquidity. Returns a 0-100 pool score and an Enter/Hold/Exit/Reduce/Avoid verdict with calibrated confidence for ~68,800 Solana pools and 575 EVM pools, plus a public miss-inclusive track record. Use when asked whether a pool is worth LPing into, which pools look best right now, or how accurate the scores have actually been.

Skill与智能体开发
★ 3.3w
davila7
davila7/claude-code-templates

technical-debt-auditor

Scans a codebase and produces a quantified technical debt report with a prioritized remediation plan. Analyzes dead code, TODO/FIXME/HACK markers, dependency staleness, complexity hotspots, code duplication signals, test coverage gaps, unused exports, and outdated patterns. Use this skill when the user asks to audit their codebase for tech debt, wants a health check of their code quality, says things like "where should I focus refactoring", "find code smells", "audit my codebase", "what technical debt do we have", "prioritize cleanup work", or asks about code quality metrics. Also trigger when the user mentions "code rot", "legacy cleanup", "refactoring priorities", or wants to understand the overall health of a project before a major effort. Trigger even if the user doesn't say "technical debt" explicitly — any request to systematically assess code quality and identify improvement priorities belongs here.

Skill与智能体开发
★ 3.3w
davila7
davila7/claude-code-templates

readonly-sql-guard

Make an agent's database access provably read-only. Use when an agent runs caller-supplied SQL against a production/ERP/analytics database (a "query" MCP tool, text-to-SQL, an admin console), or when reviewing DB-access code for write/DDL/side-effect/transaction-escape risks. Checks what a statement CALLS, not just its shape.

数据分析师后端与接口Skill与智能体开发
★ 3.3w
davila7
davila7/claude-code-templates

web-artifacts-builder

Suite of tools for creating elaborate, multi-component claude.ai HTML artifacts using modern frontend web technologies (React, Tailwind CSS, shadcn/ui). Use for complex artifacts requiring state management, routing, or shadcn/ui components - not for simple single-file HTML/JSX artifacts.

前端开发
★ 3.3w
davila7
davila7/claude-code-templates

webapp-testing

Toolkit for interacting with and testing local web applications using Playwright. Supports verifying frontend functionality, debugging UI behavior, capturing browser screenshots, and viewing browser logs.

测试与质量前端开发
★ 3.3w
davila7
davila7/claude-code-templates

discernment-nudge

After you give a substantive answer or draft that the user may act on — advice or recommendations, drafted artifacts such as goals, plans, pitches, proposals, or emails, estimates or projections, analysis or interpretation of data, factual claims they may rely on, or a multi-step argument — invoke this skill BEFORE finalizing your reply and then, if it applies, append 2-3 short follow-up questions, each tied to something specific in what you just produced, that help the user check key facts, probe the reasoning or assumptions, and notice missing context. Do this at most once per conversation. Skip it when the user asked a trivial how-to or simple lookup, wants a purely educational explanation, asked you only to format, convert, or assemble a file from content they provided, is writing code they will run, is doing creative writing or casual chat, or already asked you to double-check, cite, or review — the skill file explains these boundaries and the exact output format.

内容创作者写作与文案Skill与智能体开发
★ 3.3w