Your agency doesn’t lose money on sales.It loses money on scope.
You can win the deal at a healthy rate and still finish the project underwater. The leak isn’t your pricing — it’s the dozen small “can you just” requests between kickoff and delivery, each one too minor to invoice, collectively a real bite out of your margin.
The Agency & Studio Operations Skills Pack is seven Claude skills that sit at the boundary between what you sold and what you’re delivering — flagging scope creep, writing the SOW, pricing the change order, and sending the hard email before the free work happens. The skills check their own logic and tell you when to hold.
Scope creep is the margin leak nobody books.
You can win the deal at a healthy rate and still finish the project underwater. The leak is not in your pricing — it’s in the dozen small “can you just” requests between kickoff and delivery, each one too minor to invoice, collectively a fifth of your margin.
Project management tools don’t catch this. They track the work that already exists; they have no concept of the contract the work was supposed to live inside. By the time a task shows up “late” in your sprint board, you’ve already agreed to do it for free.
Small out-of-scope asks between kickoff and delivery, each one too minor to invoice on its own. Nobody tracks them, so the true cost stays invisible until a profitable-looking project somehow misses its margin.
The favor that's too small to bill is the favor you do for free. Multiply it across every active engagement and the uncaptured work becomes a standing line item you never get to write down.
Scope creep hides in the relationships that feel best. The friendlier the client, the harder it is to say 'that's a change order' — which is exactly why a tool that says it for you protects the margin you'd otherwise give away.
Clear about the lane. No inflated promises.
- Seven complete SKILL.md files, production-grade, with NEVER/ALWAYS guardrails.
- Template reference files you populate once: rate card, tone guide, standard terms.
- Three documented composition flows for kickoff, close, and the scope-creep close.
- A README covering all six installation paths (Claude Code, Pro, Codex, Cursor, Gemini, Copilot).
- Worked examples for each skill so you see the output before you run it.
- Built on the Agent Skills open standard — no SaaS login, no API keys, no build step.
- A project management tool — it protects the contract, it doesn't run the sprint.
- A SaaS subscription. One-time $89 with no monthly fee and no RedHub server in the loop.
- A telemetry product — your SOWs, rate cards, and client data never leave your Claude session.
- Legal advice — the skills draft documents; a lawyer reviews your contracts.
- Fake testimonials or invented case studies. None on this page, by design.
- A replacement for your account managers or your craft. It removes the operating drag so they can do the work that pays.
Seven skills. The scope boundary, end to end.
Define the boundary, watch it, price the crossings, and close the project clean. Every skill is single-purpose and composes with the others — tight skills beat one bloated mega-skill.
“is this in scope,” “client is asking for,” “out of scope”
Reads a new client request against the signed SOW and returns IN-SCOPE, OUT-OF-SCOPE, or HOLD. On HOLD it names the exact clause to re-read. Stops you from doing free work before you start it.
“write a SOW,” “statement of work,” “scope this project”
Turns a project brief into a tight statement of work with explicit deliverables, exclusions, assumptions, and acceptance criteria. The exclusions section is the load-bearing part — it is what you point to later.
“change order,” “additional work,” “out-of-scope request”
Converts an out-of-scope ask into a clean, client-ready change order priced from your rate card. Refuses to invent pricing it cannot justify. Turns scope creep into revenue instead of resentment.
“project retro,” “post-mortem,” “what went wrong”
Ingests project notes, timelines, and overruns and produces a structured retrospective: what worked, what slipped, root cause, and the one process change worth making. No blame theater — just the signal.
“status update,” “client update,” “weekly check-in”
Drafts the weekly client update in your house voice — progress, blockers, what you need from them, what's next. Calibrated to reassure without over-promising. The skill agencies use most, daily.
“QA this,” “pre-delivery check,” “ready to send”
Generates a pre-delivery QA checklist tailored to the deliverable type so nothing ships half-finished. Catches the embarrassing miss before the client does.
“reply to client,” “push back politely,” “decline this request”
Drafts the hard email — the polite, firm 'that's outside our agreement, here's how we can add it' — without torching the relationship. Pairs with the change-order writer for the close.
Look inside one skill before you buy.
This is the actual scope-creep-detector/SKILL.md that ships in the pack — abridged, not a marketing mock-up. The HOLD verdict and the NEVER block are the honesty spine in action.
---
name: scope-creep-detector
description: Detects when a client request falls outside the signed statement of work. Use when the user says “is this in scope,” “the client is asking for,” “out of scope,” or pastes a client request to evaluate against an agreement. Returns IN-SCOPE, OUT-OF-SCOPE, or HOLD with a cited reason.
---
You evaluate a client request against the project’s signed SOW and return a verdict that protects the agency’s margin.
## Your verdict — return exactly one
- IN-SCOPE — explicitly covered by a named deliverable. Quote the deliverable line.
- OUT-OF-SCOPE — not covered, or named in exclusions. Quote it. Recommend change-order-writer.
- HOLD — ambiguous, or the SOW is missing. Name the exact clause to re-read. HOLD is not a failure state; it is the correct answer whenever certainty is not available.
## NEVER
- NEVER return IN-SCOPE for a borderline request to be agreeable. Genuine ambiguity is ALWAYS HOLD. A wrong IN-SCOPE costs the agency real, unrecoverable money.
- NEVER invent or “interpret” SOW language that is not in the supplied document. If it is not written, it is not scope.
- NEVER soften an OUT-OF-SCOPE verdict because the relationship feels warm. Warm relationships are precisely where uncompensated scope creep hides.
## ALWAYS
- ALWAYS quote the specific SOW line your verdict rests on. A verdict without a citation is an opinion, not a determination.
- ALWAYS state confidence (High / Medium / Low) and what would raise it.
- ALWAYS hand off cleanly: on OUT-OF-SCOPE → change-order-writer; on a hard client reply → scope-boundary-email-responder.Three rules every output in this pack is held to.
These skills are built to tell you the truth even when it’s inconvenient. That posture is the product — the same honesty spine that runs through the whole RedHub catalog: tools that say skip, hold, or fix when the math doesn’t clear.
When a request is ambiguous against the SOW, the scope-creep detector returns HOLD and names the exact clause to re-read — it never returns IN-SCOPE just to be agreeable. A wrong yes is the verdict that costs you real, unrecoverable money. Warm relationships are precisely where uncompensated scope creep hides, so the skill refuses to soften an OUT-OF-SCOPE verdict because the client feels friendly.
The change-order writer prices every line from the rate card you supply. When a line is blank, it inserts a visible [PRICE — populate rate card] placeholder rather than guessing a number it cannot justify. The pack would rather hand you an obvious gap to fill than a confident figure you can't defend in front of a client.
The SOW generator will not write an open-ended deliverable. Every deliverable is countable, every exclusion explicit, every assumption stated. That discipline at contract time is exactly what makes the scope-creep detector able to render a clean verdict later — the two skills are designed as a closed loop.
The skills compose into the moments that matter.
A generic AI tool is a one-shot. A skill is a step in a workflow. Here is how the seven chain across the three points in a project where margin is won or lost.
Client asks for 'one more thing.' A free favor becomes a billable line in under five minutes — without a tense conversation.
Detector reads the request against the signed SOW and flags it OUT-OF-SCOPE, quoting the exact deliverable cap or exclusion it exceeds.
Change-order writer prices the extra work from your rate card and produces a clean, client-ready change order — no invented numbers.
Email responder drafts the firm-but-warm note: 'happy to help — here's how we'll handle it,' with the change order attached. The boundary holds and the relationship survives.
New project lands. You start already knowing what 'done' means — and what isn't included.
SOW generator writes the contract with countable deliverables and load-bearing exclusions. The boundary is defined before any work starts.
QA builder pre-loads the acceptance checklist from the same deliverables, so 'ready to send' is a checklist, not a feeling.
Project wraps. Internal learning and external polish from one set of notes.
Retro synthesizer pulls the lessons for your team — what worked, what slipped, root cause, the one process change worth making. Signal, not blame.
Status composer writes the final client-facing summary in your house voice — the polished wrap that sets up the renewal conversation.
This pack handles delivery ops. Two siblings handle the bookends.
Exactly what you get for $89, and what you don’t: seven skills that protect margin at the boundary — not a tool that wins the work or sets your rates. Those are the jobs of the two packs below.
Agency Operators Skills Pack
$89Handles pricing & positioning — how you win and price the work: onboarding, status reports, kickoff decks, invoice reconciliation, renewals. This pack protects the margin once you're delivering it. Own both and you cover the full engagement, win to wrap.
AI Service Pricing Kit
$79Sets the rate card the change-order writer prices against. Buy them together and the change orders read straight from your sourced pricing model — no blank [PRICE — populate rate card] placeholders left to fill.
The questions operators actually ask before deploying.
Those tools track work that already exists. They show you a task is late; they do not tell you that the task was never in the contract. These seven skills operate one layer up — at the boundary between what was sold and what is being delivered. The scope-creep detector reads a client request against the signed SOW and flags the delta before you do the work for free. Your PM tool and these skills are complementary: the tool runs the project, the skills protect the margin on it.
They are siblings that cover opposite ends of the same engagement. The Agency Operators pack handles how you win and price the work — onboarding, status reporting, kickoff decks, invoice reconciliation, renewals. This pack handles the delivery boundary — catching scope creep, writing the SOW that defines the boundary, pricing the change order, and sending the firm-but-warm email. There is a small deliberate overlap (both touch scope and status), but the lanes are distinct: one runs the agency, this one protects the margin on each project. Many agencies own both; the cross-attach section explains how they compose.
All seven follow the Agent Skills open standard (a folder with a SKILL.md file). They install in Claude Code, Claude Pro and Team via the Skills panel, Codex CLI, Cursor, Gemini CLI, and GitHub Copilot's agent mode. The pack ships with a README covering all installation paths. No code changes, no API keys, no build step — you drop the folders in and Claude loads them automatically when the task matches.
Every skill in this pack has explicit guardrails. The scope-creep detector will not silently approve a borderline request — when the delta against the SOW is ambiguous, it returns a HOLD verdict and tells you exactly which contract clause to re-read before responding to the client. The change-order writer refuses to generate pricing it cannot justify from the rate card you supply. This is the same honesty spine that runs through the whole RedHub catalog: tools that say skip, hold, or fix when the math does not clear, instead of producing confident output that costs you money.
Solo operators are the people who bleed the most on scope, because there is no account manager standing between you and the client's 'quick favor.' These skills give a one-person shop the deal-protection instincts of a larger agency's ops department. The SOW generator and change-order writer in particular pay for the pack the first time they turn a 'can you just' into a billable line item.
No. Skills are instruction files — they run entirely inside whatever Claude environment you already use. There is no RedHub server in the loop, no telemetry, and no external API. Your SOWs, rate cards, and client communications never leave your Claude session. You can audit every file before installing; they are plain markdown.
Yes, and you should. Each skill reads from reference files you control — your rate card, your tone guide, your standard terms. The change-order writer pulls pricing from your rate card, not a generic one. The status-update composer matches your house voice. The pack ships with template reference files you populate once; after that, every output is yours, not a template's.
Thirty days, no questions. If the pack does not save you billable time or protect a single scope boundary in the first 30 days, email us and we refund in full. Digital products are hard to return, so the only thing that makes this fair is standing behind it — which we do.
Stop doing the next favor for free.
Seven skills, one afternoon to install, and the first scope boundary they protect pays for the pack. Installs in Claude Code, Pro, Codex, and Cursor. Yours forever. Thirty-day refund if it doesn’t earn its price.
Sold by RedHub AI LLC · Secured by Stripe · redhub.ai