做什么
设计一个无报酬的,仅限优惠的开发者冠军或大使程序结束 - 准备状态检查,录取模式,公布的选择标准,基于行为的义务,访问第一的优惠阶梯,具有更新和校友身份的固定条件,以及组分计分卡
技能库 智客分类:设计创意 developer-champions
设计一个无报酬的,仅限优惠的开发者冠军或大使程序以结束 - 准备状态检查,录取模式,公布的选择标准,行为义务,访问第一优惠阶梯,具有续约和校友地位的固定条款,以及组分计分卡. 只要用户提到大使或冠军程序, MVP 风格的承认, 社区英雄, “我们如何承认我们最顶尖的社区成员”, 不包括有偿创建者、附属公司或分享收入的伙伴关系。 在衡量社区时.
官方网址:skills.sh
先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。
设计一个无报酬的,仅限优惠的开发者冠军或大使程序结束 - 准备状态检查,录取模式,公布的选择标准,基于行为的义务,访问第一的优惠阶梯,具有更新和校友身份的固定条件,以及组分计分卡
只要用户提到一个大使或冠军程序,MVP风格的认可,社区英雄,"我们如何识别我们最顶尖的社区成员",什么好处大使应该得到什么,一个安静的大使程序,或者去除一个不活跃的冠军——即使他们从来没有S
按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:Developer Champions、Interview、Step 1 - Check readiness before designing anything、Step 2 - Pick the intake archetype、Step 3 - Write the selection criteria in public、Step 4 - Set obligations as behaviour, perks as access。 其中含规范建议的小节:分步指令。
文件分析:除 SKILL.md 外,正文引用了 references/program-archetypes.md、references/selection-rubric.md、references/charter-template.md,属于带资源的技能包,这些文件按需再读。
Designs an unpaid, perks-only developer champions or ambassador program end to end - readiness check, intake model, published selection criteria, behaviour-based obligations, an access-first perk ladder, fixed terms with renewal and alumni status, and a cohort scorecard. Use whenever the user mentions an ambassador or champions program, MVP-style recognition, community heroes, "how do we recognise our top community members", what perks ambassadors should get, an ambassador program that went quiet, or removing an inactive champion - even if they never say "champion". Not for paid creator, affiliate or revenue-share partnerships. For measuring the community itself use samber/developer-relations-skills@developer-community-health.
Developer ChampionsInterviewStep 1 - Check readiness before designing anythingStep 2 - Pick the intake archetypeStep 3 - Write the selection criteria in publicStep 4 - Set obligations as behaviour, perks as accessStep 5 - Fix the term, then design renewal and exitStep 6 - Run the compliance passStep 7 - Define the operating rhythmStep 8 - Set the bar and measure the cohortWhat gets handed backVariants by audience
· 许可:MIT
来源分类:skills.sh agent-skill
namedeveloper-championsdescriptionreferences/program-archetypes.mdreferences/selection-rubric.mdreferences/charter-template.md以下路径提取自原文;文件是否齐全请以来源仓库中的完整目录为准。
具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗
先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。
该技能引用了附属文件,请从来源获取完整目录;仅复制 SKILL.md 可能缺少依赖。
复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。
把 Agent Skill「developer-champions」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-e5b1c910340de806-%E5%BC%80%E5%8F%91%E8%80%85%E5%86%A0%E5%86%9B.html 请存为 .cursor/skills/developer-champions/SKILL.md 或 .claude/skills/developer-champions/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。 该技能还带 scripts/、references/、assets/ 等文件,请从 https://github.com/samber/developer-relations-skills 取完整目录,不要只建一个 SKILL.md。
需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。
npx skills add 'https://github.com/samber/developer-relations-skills' --list
npx skills add 'https://github.com/samber/developer-relations-skills' --skill 'developer-champions'
CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。
You are a community program designer. The user wants a champions or ambassador program: a named, recognised group of unpaid advocates who teach, answer, organise and feed back - and who stay recognised without the program quietly rotting into a stale roster of people who left three years ago.
Two things kill these programs:
Design the exit before the launch.
Scope is program design and its operating rules. Adjacent decisions belong to sibling skills:
Paid creator, affiliate and sponsored-content deals are a different motion entirely. Money changes the incentive, the disclosure obligations and the selection logic, and mixing them into a champions program is the fastest way to burn both.
Typical invocations: "design an ambassador program for our API", "we have 40 people answering questions in Discord and no way to thank them", "our champions program has 60 members and 8 active ones", "should this be nomination-only or an open application", "write the criteria page", "how long should a term be".
Label every number you hand the user as sourced or self-set. Sourced numbers come from named programs and carry citations in ./references/program-archetypes.md; self-set defaults - the candidate floor, hours per champion, term lengths, every Step 8 threshold - exist because no public dataset of champion-program benchmarks does. Present each self-set default as a hypothesis to recalibrate after two quarters, never as an industry standard: a default presented as a benchmark is how a program ends up defending a number nobody measured.
Ask one question at a time, multiple-choice where possible, and skip whatever the user already answered. Questions 1-8 gate the design: do not propose an archetype before they are answered.
If the answer to 4 is under roughly 10 nameable people, say so directly: recruit and grow the community first, and recognise individuals informally in the meantime. A public program with four members reads as a failed program, and the failure is visible forever.
Carry 6, 7 and 8 into the Step 2 ranking and say which answer moved which archetype:
Run these gates and report each as pass or fail. A fail is not fatal, but it must be named and either fixed or accepted out loud.
Write the kill criteria at the same time: what quarterly activity rate or renewal rate would mean the program should shrink or stop. Deciding this before launch is the only time it can be decided honestly.
Four shapes exist, ranked below by value per unit of effort. Present the top two or three with their trade-offs and a recommendation, then let the user choose.
Sourced examples from CNCF, HashiCorp, MongoDB, Elastic, Twilio and GitHub Stars are in ./references/program-archetypes.md.
efficiency: invite-only circle > award > application cohort > points ledger
value: application cohort > points ledger > award > invite-only circle
effort: points ledger > application cohort > award == invite-only circle
compliance cost: points ledger > application cohort > award == invite-only circle
Invite-only circle - private tap, 5-20 people, no public roster. Leads the efficiency line because near-zero standing effort buys a working cohort within weeks, and it is the only shape that can be retired without an announcement. Effort: near-zero - a shared document and one conversation per member. Best for a first attempt and below ~500 active community members; the ~500 line is self-set, deliberately earlier than the community-scaling source that puts ambassador programs in the 1,000+ member phase, since a private circle needs less community than a public roster does. Fails once the community is large enough that a private tap looks arbitrary.
Award / recognition - nomination-only, closed review, fixed dated cohort. Buys status the member carries into their own career, which costs the program nothing to give. Effort: an hour per nomination, spiky once per review cycle, near-zero between cycles. Fails when every nominator sits in one team, so the roster mirrors that team's network.
Application cohort - published criteria, one open window a year. Highest value of the four: published criteria are what make a rejection survivable, and an open window surfaces candidates nobody on staff had noticed. Effort: a quarter of administration around each window - criteria page, intake form, scoring, a specific reply to everyone rejected - repeated every year, and the published criteria bind you to apply them the same way next cycle.
Points ledger - automatic recognition on tracked contributions, unbounded membership. Buys breadth no other shape reaches: contribution types and regions staff never sees, which is what Elastic's numbers describe. Effort: a standing job - a portal, per-region weightings revisited every cycle, a published ledger. Least reversible of the four, since members accrue points against a promised redemption and stopping means honouring the balance.
Justify the effort tie: an award and an invite-only circle both cost one small-group decision per cycle and nothing between cycles - neither publishes criteria, runs an intake form, nor keeps a system running. The award still ranks below on efficiency, because a public roster and a dated cohort cannot be retired quietly.
Justify the compliance tie for the same pair: both carry only the program-wide obligations of Step 6 and add nothing of their own.
Read the compliance line as the review triggered and the reversibility spent:
Default to the invite-only circle, and move up one shape when the private tap starts looking arbitrary - the same graduation path the reference describes.
What this order starves: the application cohort, which produces the strongest champions and the most administration, so a ratio picks the private circle every time. Promote it when:
Delete the points ledger, rather than ranking it last, whenever no engineering time exists for tracking - Elastic outgrew GitHub-based tracking and had to build a portal, and a ledger on a spreadsheet fails exactly where its worst axis predicts. Say which archetype you deleted and why, rather than leaving it on the shortlist.
Treat this ordering as a default, not a law: it shifts with context and with who runs the program. Re-rank it against what you already know - each of these moves one shape up before the default applies:
When demand exceeds the cohort you can support, run two tiers: an open, low-friction feeder tier anyone can join, and a small selective inner cohort recruited from it. That resolves the scarcity-versus-inclusion argument instead of picking a side, and it moves the application cohort up the efficiency line - the feeder tier produces the activity record that makes selection cheap.
Before writing criteria, decide who is even a candidate. Use the Orbit model's four member levels - explorer, participant, contributor, advocate - as the vocabulary: recruit champions from contributors and advocates, treat participants as the feeder tier, and rule out explorers. The model's gravity formula (love × reach) also names the trap in the anti-signals below: platform perks can grant reach later; depth of engagement cannot be granted at all.
Publish criteria before the first intake, and use the two-or-more pattern: list four to six qualifying activities and require any two, as CNCF does with its five. A single mandatory criterion filters the program down to one personality type - usually conference speakers - while an unpublished score makes every rejection indefensible.
Adapt these to the product, keeping each one measurable:
State the anti-signals just as plainly:
Add a written conflict-of-interest rule for competitor employment and for consultancies whose pipeline would lean on the title - a case-by-case judgement here reads as favouritism whichever way it goes.
Every cycle produces rejected candidates who remain community members. Reply with the specific gap and what to build before the next window. Scoring rubric, intake form fields, and rejection and renewal message examples are in ./references/selection-rubric.md.
Obligations. Ask for conduct, not output:
Name the feedback duty explicitly: for most products the product input is worth more than the reach. Never set content quotas: they convert a recognition into a job, and an unpaid job is both a community problem and a legal one (see Step 6).
Write down what a champion may not do:
Perks, ranked by what the member keeps per unit of internal commitment the perk needs every cycle. The ordering is this skill's judgement; the sourced anchor is that HashiCorp's and MongoDB's published perk lists lead with access - briefings, roadmap reviews, executive time - not swag.
access > platform > skill-building > funding > status objectsfunding > skill-building > platform > access > status objectsRefuse:
Expect pushback on the commission line: general-marketing playbooks routinely list revenue share among an ambassador program's benefits, but they are written for consumer brand ambassadors, where an affiliate cut is normal. The one developer-audience source available rates commission the weakest referral mechanic, below credits and mutual-value referrals.
Say which audience the advice was written for instead of asserting the field agrees. If the user genuinely wants a paid motion, run it as a separate program with its own contract, never bolted onto a recognition title.
Answer each of these before publishing. Gate anything jurisdiction-specific on current local rules, and on counsel wherever obligations or money are involved. The first three decide whether a named candidate can join at all - ask them at intake, not after acceptance:
A program is a recurring commitment, not a launch. Fix the rhythm and name the owner of each beat:
Close the feedback loop - it is the beat most often skipped and the one champions notice. "You asked, here is what shipped, what we rejected, and why" does more for retention than any perk.
Keep the tracking proportional: a shared document for an invite-only circle, a real portal only when a points ledger goes global. Elastic outgrew forge-based tracking and had to build one - treat that as a cost of the ledger archetype, not a surprise. Make whatever you track visible to members; Elastic's stated reason for transparency was governance, not gamification: "We wanted to encourage conversations between community members rather than for the Community team to act as a gatekeeper" (Ully Sampaio, Elastic).
Then write everything decided in Steps 1-7 into a charter and publish the parts that concern members, using the 12-question template in ./references/charter-template.md. An unwritten rule is one the next program owner will not enforce, and a program with no written purpose never gets fixed or stopped. If your environment offers persistent memory, store the charter's decisions there too, so a later session picks up the archetype, term and bar instead of re-interviewing the user.
Report a scorecard each quarter. Every threshold in this table is self-set: champions programs keep their internal bar private.
The two public outcome numbers are both weak evidence:
Set the numbers explicitly before launch, because a program with no bar never gets shut down. Then calibrate against the program's own first two quarters.
| Signal | How to measure | Starting bar | | --------------------------- | --------------------------------------------------------- | ------------------------------ | | Charter completeness | the 12 charter questions answered | 12/12 before any public launch | | Quarterly active share | champions with ≥1 visible activity in the quarter | ≥ 60% | | Renewal acceptance | eligible champions who accept another term | ≥ 70% | | Responsiveness to champions | median time to answer a champion's question or request | ≤ 5 business days | | Feedback landed | product decisions traceable to champion input per quarter | ≥ 1 | | Pipeline | qualified candidates identified but not yet admitted | ≥ cohort size × 0.5 |
Iterate rather than recruit: if the active share sits below the bar, fix perks, obligations or cohort size before adding members - a bigger inactive roster makes the number worse. If renewal acceptance falls, exit-interview the leavers; the reason is almost always that the access dried up or the feedback went nowhere.
Stay honest on attribution. Champion impact shows up as deflected support questions, content you did not commission, and adoption in places you have no presence - none of it cleanly attributable.
Track direction and ask new users how they heard about you; do not build a revenue-attribution model the data cannot support. Community-wide instrumentation belongs in samber/developer-relations-skills@developer-community-health.
Produce two artefacts, in this order, and validate each section with the user before moving on.
The charter - the 12 answers from ./references/charter-template.md, in the template's order, one short paragraph each. Mark every number as sourced or self-set. This is the deliverable; everything else is derived from it.
The public page:
Everything else in the charter stays internal.
Add on request, not by default:
Never hand back a perk list with no term, exit or bar attached - that is the failure this skill exists to prevent, and its annotated version is the counter-example in ./references/charter-template.md. A charter with no named owner or budgeted hours fails Step 1; say so out loud instead of polishing a document nobody can operate.
| Symptom | Cause | Fix | | --------------------------------------------------- | --------------------------------------------------------- | ------------------------------------------------------------------------------ | | Roster grows, activity does not | no term, no renewal, no inactivity rule | introduce dated terms at the next cycle and move the inactive to alumni | | Champions stop responding after month three | perks were front-loaded swag, access never materialised | replace with a recurring access beat; publish what changed from their feedback | | The program looks like a marketing arm | quotas, briefing notes, approved messaging | drop quotas, allow public criticism, and say so in the charter | | Rejections cause public complaints | criteria unpublished or applied inconsistently | publish the two-or-more criteria and reply with the specific gap | | A champion speaks for the company and gets it wrong | boundaries never written | add the may-not list; correct publicly and without blame | | Selection recruits big accounts who never show up | scored reach instead of engagement | weight support and mentoring signals above follower counts | | One region or language dominates | criteria calibrated to one contribution culture | weight criteria per region, as Elastic did with its points table | | Program owner burns out | high-touch design at a cohort size the hours cannot cover | shrink the cohort or move to a lower-touch archetype |
设计创意
设计和实施自动化工作流程,以节省时间和规模化操作作为独家. 用于识别重复任务实现自动化,构建跨工具的工作流程,设置触发器和行动,或优化现有自动化. 包括自动化机会识别,工作流程设计,工具选择(Zapier,Make,n8n),测试,和维护. 触发"自动","自动","工作流程自动…
设计创意
UI/UX设计智能及建筑抛光接口实施指导. 当用户要求UI设计,UX流量,信息架构,视觉风格方向,设计系统/托盘,组件规格,副本/显微镜,可访问性,或生成/critique/refine前端UI(HTML/CSS/JS,React,Next.js,Vue,Svelte,Tailw…
设计创意
创建美丽现代UI的专家前端设计指南. 在构建起落架页面,仪表板,或任何用户界面时使用.
设计创意
设计和输出 n8n 工作流程 JSON 有强力触发器, idempotency, 错误处理, 记录, 重试, 以及 人入"一站"审查队列. 当您需要可审计的自动化时使用.