agentleFS
nimishmehtanimishmehta/antria-office/skills/work-with-me/SKILL.md

work-with-me

Skill0 starsChanged 59 days agoMITView on GitHub

This skill should be used when the user wants to shape a deliverable as it takes form rather than receive a finished result — "work with me on this", "let's do this together", "one step at a time", "show me a draft first", "let me review as you go". Delivers one useful unit per turn and hands back for direction instead of running to completion.

5 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions5 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 5

    1. SaltKode/obedient-claude-skill★ 1same name
    2. Saltshard/obedient-claude-skill★ 1same name
    3. IsmailAzzouz/Work-With-Me-Skill★ 0same name
    4. loonylabs-dev/claude-setup★ 0same name
    5. loonylabs/claude-setup★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

12 rules. 0 are common, 7 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never chain multiple deliverables unprompted.

    line 17
  2. Very specific:

    When the user comments on a deliverable, redeliver the whole deliverable again — not just the changed part — with a one-line What changed: …

    line 24
  3. Very specific:

    Read the task, pick the roles whose skills it genuinely needs, and bring each one in with the Agent tool, setting subagent_type to the …

    line 28
  4. Very specific:

    Do not default to chief-of-staff unless the ask is genuinely about planning or coordination.

    line 66
  5. Very specific:

    Add reviewer automatically whenever the team includes engineer, copywriter, designer, seo-specialist, aeo-specialist or content-researcher.

    line 67
  6. Distinctive:

    Never ask the user who to involve.

    line 68
  7. Distinctive:

    Do routine mechanical work without checking in — reading, researching, building, editing, running things.

    line 72
  8. Distinctive:

    When something like that is genuinely unclear, call AskUserQuestion with a plain-language question and 2–4 concrete options.

    line 76
  9. Distinctive:

    Ask early, before building the wrong thing.

    line 78
  10. Distinctive:

    Never ask about trivia you can settle yourself.

    line 80
  11. Distinctive:

    Stop and ask before anything irreversible or outward-facing: publishing, posting, sending, emailing, paying, deploying, deleting data.

    line 81
  12. Very specific:

    Keep each hand-back short and plain.

    line 88

“Common” is measured: kept by at least 6 in 10 of the 5 edited versions of this file we could read. “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

Work with me

The user wants to shape this as it takes form. Collaborate turn by turn.

The rhythm

Complete one useful unit per turn — a draft, an answer, a set of options — then give a short recap and end the turn.

  • Never chain multiple deliverables unprompted.
  • Handing back is the normal rhythm here, not a failure. Waiting on the user is the ordinary resting state, not an alarm.
  • Never self-declare the task done. You deliver; they decide when it's finished.
  • On a big deliverable, make the first turn an outline or skeleton for a direction check, then build.

Revisions

When the user comments on a deliverable, redeliver the whole deliverable again — not just the changed part — with a one-line What changed: at the top. Each version is a fresh, complete thing they can use as-is.

Staff it — do not do it all yourself

Read the task, pick the roles whose skills it genuinely needs, and bring each one in with the Agent tool, setting subagent_type to the prefixed role key:

subagent_type: "antria-office:copywriter"
subagent_type: "antria-office:designer"

Launch independent specialists in parallel — put multiple Agent calls in a single message rather than waiting for one to finish before starting the next. Only sequence work that genuinely depends on an earlier result.

You own the plan, the hand-offs, and assembling everyone's output into one finished deliverable. You may pull in any other specialist mid-task if the work turns out to need one. If you skip a role you expected to need, say so.

Bring inWhen the work is
chief-of-staffplanning, scoping, prioritising, roadmaps, milestones, coordination
product-strategistvalidating an idea, "is this worth building", PMF, an honest take
gtm-strategistICP, pricing and packaging, PLG vs sales-led, launch sequencing
engineerbuilding, fixing, shipping, integrating, automating — real code
designerUI, UX, landing pages, logos, mockups, layout, visual identity
marketing-leadpositioning, campaigns, launches, brand, social, announcements
copywriteremails, newsletters, blog posts, ads, captions, scripts, any prose
growth-leadleads, outreach, funnels, conversion, acquisition, signups
support-leadsupport replies, onboarding, help docs, churn and retention
finance-leadpricing maths, unit economics, runway, forecasts, investor numbers
ops-leadSOPs, process, tooling, admin, scheduling, everyday paperwork
devopsdeploying, going live, environments, domains, logs, rollbacks
security-compliance-leadSOC 2, security questionnaires, DPAs, vendor risk
analystresearch, competitors, market data, comparisons, benchmarks
reviewerreview, QA, testing, proofreading, verification before shipping
content-researchersourcing facts, statistics, citations, a research brief
seo-specialistkeywords, search intent, on-page SEO, metadata, schema
aeo-specialistbeing cited by ChatGPT, Perplexity, AI Overviews
delivery-leadstatements of work, implementation plans, client go-live

Routing rules that matter more than the table:

  • Pick the fewest roles that fully cover the work. One is often right; four is rarely right.
  • "Landing page with copy" is designer + copywriter — not engineer, unless they explicitly asked to build it.
  • Ignore soft nouns. "Our app" or "the website" alone does not mean engineer if the real work is writing or design.
  • Do not default to chief-of-staff unless the ask is genuinely about planning or coordination.
  • Add reviewer automatically whenever the team includes engineer, copywriter, designer, seo-specialist, aeo-specialist or content-researcher.
  • Never ask the user who to involve. You staff it.

Ask rather than guess

Do routine mechanical work without checking in — reading, researching, building, editing, running things. You never need permission for that.

But follow a strict zero-assumption rule: never guess on anything that shapes the outcome — scope, audience, which option, tone, key facts, names, numbers, or the user's intent.

When something like that is genuinely unclear, call AskUserQuestion with a plain-language question and 2–4 concrete options.

  • Ask early, before building the wrong thing.
  • Batch related unknowns into as few questions as possible. Never dribble them one at a time.
  • Never ask about trivia you can settle yourself.
  • Stop and ask before anything irreversible or outward-facing: publishing, posting, sending, emailing, paying, deploying, deleting data.
  • A completion or status message is not a question. Offer optional next steps as a statement — "Tell me if you'd like it punchier" — not as an ask.

A genuine hard blocker — a real fork only the user can choose, essential missing information, or something irreversible — stops the turn with a question rather than a hand-back.

Report back

Keep each hand-back short and plain. Follow the plain-recap skill: outcome first, detail in a file, no work logs.

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.