curviate
Entry point for driving LinkedIn work through the Curviate CLI: which skill answers which task, how to resolve the `curviate` binary, the retrieval-mode concept and the flags only four commands accept, and the rule that an agent branches on the exit code rather than on the message text. Use whenever a task involves LinkedIn profiles, companies, search, messages, posts, comments, reactions, connections, follows, job postings, Sales Navigator or Recruiter, or whenever a `curviate` command is about to run and it is not obvious which area covers it.
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 3.4w
- Forks
- 4.0k
01 前言 Front Matter
- name
- curviate
- license
- MIT
- metadata
- { "tags": [ "LinkedIn", "CLI", "Agents", "Sales", "Recruiting", "Outreach" ], "author": "Curviate", "version": "0.1.0", "repository": "https://github.com/Curviate/curviate-plugin" }
- description
- Entry point for driving LinkedIn work through the Curviate CLI: which skill answers which task, how to resolve the `curviate` binary, the retrieval-mode concept and the flags only four commands accept, and the rule that an agent branches on the exit code rather than on the message text. Use whenever a task involves LinkedIn profiles, companies, search, messages, posts, comments, reactions, connections, follows, job postings, Sales Navigator or Recruiter, or whenever a `curviate` command is about to run and it is not obvious which area covers it.
02 资源文件 3
- 参考文档AGENT-BRIEF.md7.8 KB
- 参考文档agents/openai.yaml135 B
- 参考文档OUT-OF-SCOPE.md4.6 KB
03 SKILL.md 正文
Curviate: start here
Curviate is a LinkedIn API for agents. The curviate CLI is its command surface, and this skill is
the dispatcher: it says which skill to read, and carries the few rules that apply everywhere. It
carries no command syntax: every command, flag and worked example lives in the area skill that owns
it.
Established against CLI 0.33.0.
Which skill answers this
| Task | Skill |
|---|---|
| First run on a new machine: install, authenticate, connect an account, prove it works. | curviate-quickstart |
| A member profile or a company page. Resolving a human search term into the filter id a search needs. Sessions, stored profiles and connected accounts. | curviate-profile |
| Finding people, companies, posts, jobs, services or groups. Running a pasted LinkedIn search URL. | curviate-search |
| Reading conversations, sending and managing messages, InMail, company-page inboxes. | curviate-inbox |
| Publishing posts, commenting, reacting, the home feed, notifications. | curviate-engage |
| Connection invitations, follows, relations, followers. | curviate-network |
| Job postings: creating, publishing, budgets, applicants. | curviate-jobs |
| Sales Navigator and Recruiter. | curviate-premium |
When a task spans two areas (find someone, then message them), read both. They are self-sufficient and repeat what they need to; nothing is hidden behind a third skill.
Resolve the binary before anything else
curviate --version # needs Node 18 or newer
If that fails, install it: npm install -g @curviate/cli. Do not substitute a raw HTTP call for a
command you could not find. The command surface is the contract, and hand-rolled requests miss the
projections, the retry contract and the exit-code mapping that every one of these skills is written
against.
Credentials resolve flag > environment > stored profile. For a scripted run prefer
CURVIATE_API_KEY in the environment or a stored profile: a key passed as a flag is visible to other
processes through ps and is saved in shell history.
Branch on the exit code, never on the message text
Every command exits with a code that says what to do next, and under --json an error prints
{"error": {"code", "message", …}}. The code is the contract. The wording is not: it is written
for a human reading a terminal and it changes between releases.
Two consequences worth internalising before the first failure:
- A single exit code can carry several distinct causes, so read
error.codebefore deciding on a remedy. Exit5alone means three unrelated things whose fixes have nothing in common. - Retrying is right for some codes and actively wrong for others. A rate limit tells you how long to wait; a refusal by one of your own ceilings will not clear on a retry at all. Each area skill ends with the exit codes reachable from its commands and the action each implies. Read that table rather than guessing from the message.
Retrieval mode, and the flag of the same name that is not it
Some reads can be answered from a stored copy instead of a live call to LinkedIn, and --mode /
--max-age choose between them. Two rules matter everywhere:
- Exactly four commands accept those flags:
profile me,profile <id>,inbox getandinbox messages. The mechanics (what each mode does, what a store miss returns, and whycache_onlyand--max-agecannot be combined) are incurviate-profileandcurviate-inbox. - Every other command refuses them outright rather than ignoring them:
unknown flag `--mode`at exit2, before any network call. That is the good failure. A retrieval-mode habit carried onto a search fails loudly instead of quietly serving you a freshness you did not ask for. The two exceptions below are not retrieval commands either: they refuse retrieval values.
Two commands take a --mode that is a different flag wearing the same name, and both spend real
money: job publish and recruiter job publish, each --mode FREE|PROMOTED|PROMOTED_PLUS. They
select how a posting is published. No retrieval value is valid there, and on both the flag is
required rather than optional, so a retrieval habit is refused at exit 2 rather than silently
promoting a posting. See curviate-jobs and curviate-premium.
Reading a response
- Pass
--jsonon anything you parse. It is already the default when stdout is not a terminal; ask for it explicitly when a human might also be watching. - An empty field in a default response is not proof the data does not exist. Responses are
projected, and several fields are withheld until
--verboseasks for them, including a few that read exactly like "this member published nothing". Each area skill names its own. --previewbefore a write renders the resolved request without sending it. On a read command it is refused at exit2.
04 同仓库更多 Skills
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.
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.
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.
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.
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.
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.