quote-build — Claude Code skill: quoting and installation
The quote-build Claude Code skill builds make-to-order quotes from intake forms, past jobs, and cost history. It runs on requests to price a job or revise a quote after a customer counters.
Apache-2.0 ★ 47,914 updated 2026-10-24
npx skills add anthropics/claude-plugins-official --skill quote-build -a claude-codeWhat quote-build does
The skill prepares a customer quote and a separate cost-basis sheet. It compares the job with prior work, breaks machine hours down by operation, and keeps setup time apart from run time. Material cost follows actual purchase prices, margin gets its own line, and setup cost is recalculated at each quantity break.
Before committing to a delivery date, the skill checks the shop schedule. When a customer pushes back on price, it revises the existing quote and records what changed and why. Connectors are optional: the skill can work from an intake form and pasted cost figures; connected data sources include an ERP, Microsoft 365 documents, a shop-load export, and Autodesk Fusion for CAM-derived hours.
What it helps with
- price a make-to-order job from its intake form and prior work
- trace material cost to an actual purchase price
- separate setup hours from run hours
- check a promised date against shop capacity
- revise a quote after a customer counters on price
Example prompts for the agent
- Build a quote from this customer intake form.
- Price this job using the prior work and cost figures.
- The customer came back on price; revise the quote and record the changes.
How to install quote-build
1Skills CLI (Vercel)
npx skills add anthropics/claude-plugins-official --skill quote-build -a claude-code2Manually: put the files in place
~/.claude/skills/quote-build/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/quote-build/
1Skills CLI (Vercel)
npx skills add anthropics/claude-plugins-official --skill quote-build -a codex2Manually: put the files in place
.agents/skills/quote-build/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add anthropics/claude-plugins-official --skill quote-build -a cursor2Manually: put the files in place
.cursor/skills/quote-build/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install anthropics/claude-plugins-official plugins/claude-for-manufacturing/skills/quote-build2Manually: put the files in place
.github/skills/quote-build/SKILL.md1Skills CLI (Vercel)
npx skills add anthropics/claude-plugins-official --skill quote-build -a gemini-cli2Platform CLI
gemini skills install https://github.com/anthropics/claude-plugins-official --path plugins/claude-for-manufacturing/skills/quote-buildManually: put the files in place
.devin/skills/quote-build/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about quote-build
How do I install quote-build in Claude Code?
Run npx skills add anthropics/claude-plugins-official --skill quote-build -a claude-code. The skill file is installed at ~/.claude/skills/quote-build/SKILL.md.
What does quote-build do?
It prepares a make-to-order quote with a separate breakdown of its cost basis. It also checks the promised date against shop load and records revisions after price negotiations.
What does quote-build need to run?
An intake form and cost figures are enough; connectors are optional. With integrations, it can read NetSuite or Infor M3, Microsoft 365 documents, a shop-planning export, and CAM hours from Autodesk Fusion.
Source text
SKILL.md · Apache-2.0
--- name: quote-build description: "Produces the make-to-order quote from the intake form, comparable prior jobs and cost history, with every number traced to where it came from — material at the price actually paid on a named purchase order, machine hours per operation from a labelled ladder (current routing, closed-job actuals, the confirmed design, first principles) with setup and run kept apart, setup re-amortized at each quantity break, and margin on its own line. Checks the promised date against real shop load before committing to it. Revises in place when a customer counters, preserving what changed and why. Use when someone says \"build the quote,\" \"price this job,\" \"what should we charge,\" \"quote this RFQ,\" or \"the customer came back on price.\"" compatibility: "Requires none. Reads NetSuite or Infor M3 (or any ERP mapped into the dataset) for standard cost, actual cost on closed work orders and purchase prices, Microsoft 365 for prior quotes and templates, a planning export for shop load, and Autodesk Fusion for CAM-derived hours where connected. Runs on the intake form and pasted cost figures with no connector. Reads rates from an MFG_* variable or the plant profile, never asking each run. The quote uses the plant's own template." --- # Quote Build Pricing a make-to-order job means building up a cost and then applying commercial judgment. Every job is priced fresh, which is why quoting is skilled, slow, expert work — and why the estimator is usually the bottleneck. **The cost basis is shown, always.** A quote that says *$14,720* is useless. One that says *$14,720, of which $6,100 is material at the price we paid on PO-8841 in June, plus 8% scrap* is defensible. **Quick start** ``` → Load the intake form. Lines with no quantity or no material are skipped; quote the rest and list what's blocked → Find comparables; show what made each one comparable AND how this job differs → Flag capability BEFORE pricing. Then hours per op from the ladder: routing → actuals → confirmed design → first principles → Setup and run separate at every rung. A superseded routing is shown, never priced. probable_standard never outranks rung 3 → Build up the cost: material, labor, machine, setup, outside processing, tooling → Resolve rates: MFG_* variable, then the plant profile. Never asked in the run, never invented → Re-amortize setup at every quantity break → Margin on its own line. Never buried → Check the promised date against real shop load. Never quote a date blind → Assemble the customer quote AND the internal cost-basis sheet. No hours, ranges, rates, paths or basis labels in the customer's → On a counter: revise in place with a change record. Never a second quote, never reopen CAD ``` --- ## Inputs **First, find the plant setup:** from the task's own folder (never `cd` into the plugin), run `python3 -I <this skill's folder>/../../shared/scripts/plant_setup.py find`, do its `setup_page_first` step, then read the profile at `profile_path` and any rate at `rates_path`, per `../../shared/replies.md`. | Input | Source | Form | |---|---|---| | The intake form | `customer-inquiry-intake` | `../../shared/contracts/intake-form.md` | | Consolidated dataset, including `routings` and `operation_actuals` (v4) | `data-consolidate` | `../../shared/contracts/consolidated-dataset.md` | | Geometry estimate, per line | `rfq-to-quote` step 7, or typed by the estimator | `../../shared/contracts/geometry-estimate.md` | | Machine classes, cutting data, variance threshold | `### Machining data` in the `## Plant context` block | Config | | Standard cost and item master | The ERP (NetSuite or Infor M3), each cost read from the field the profile's cost map names (`../onboarding/reference/erp-scope.md`); a field equal to the sell price is never a cost. Infor M3 gives no standard cost: last purchase price first, per `../../shared/erp-connectors.md` | Connector | | Actual cost on prior jobs | `production_work_orders` actual-cost columns in the dataset (v6): NetSuite closed work orders, Infor M3 work order operations, or any ERP `custom-connector` mapped | Dataset | | Recent material purchase prices | ERP purchase orders | Connector | | Awarded supplier bid for an outside process, with its selection record | `supplier-quote-request`, when this quote started the sourcing event | `../../shared/contracts/supplier-bid.md` | | Prior quotes for this customer or part | M365 documents, NetSuite. Infor M3: documents only | Documents and connector | | Labor, burden and margin rates | `MFG_*` variable, else the plant profile — see below | Config | | Shop schedule and load | Planning export to the M365 folder | CSV/XLSX | **Rates are resolved, not asked for each run**, in the order `../../shared/secret-config.md`, *Resolving a rate*, sets, and no other: 1. `MFG_LABOR_RATE`, `MFG_BURDEN_RATE`, `MFG_MARGIN_RULES`, checked by name, never printed. These win; a stored value for the same rate is **never printed anywhere**, only named as also held. 2. A **confirmed** rate in the `## Plant rates` file at `rates_path` (an older profile: its `### Commercial rates`). Price on it with no label. 3. A **placeholder** in the same place. Price on it and label it, below. 4. Nothing. **Block the lines it feeds**, named; the rest is still produced. Shown on the internal cost-basis sheet, never on the customer-facing quote. **A rate typed into this run is not used**: say it is not on file, do not repeat the number, and offer `onboarding`'s "update our labour rate", which saves it with a name and date for every quote after this one. --- ## Workflow ### Step 1 — Load the intake form **Refuse to proceed on line items with no quantity or no material.** Quote the rest and list what is blocked and why. A 40-line RFQ with one underspecified item produces 39 prices and one question, not silence. **Read the commercial terms** — payment, freight, packaging, liability. Where the intake form states one, carry it into the quote as given. Where one is null, quote against the plant's standard for that term, name the assumption on the exclusions list, and keep the missing-register row — do not fill it in silently. Where `customer-inquiry-intake` flagged a stated term as clashing with the plant's standard, carry that flag onto the exclusions list too; it does not resolve itself by arriving here. **Block on missing provenance.** Where `field_sources` is absent from the intake form, block — without provenance the estimator cannot verify a number against the package, and a quote built on an unsourced field is a quote no one can defend. **Where `quantities` has only one entry but the customer's ambiguous register shows breaks were requested**, quote what is there and state plainly that the breaks were not captured — never invent the missing quantity points. **Where `ambiguous_register` carries an entry on a field the quote depends on**, quote against one value, name which, and state the other explicitly rather than silently picking one — the same discipline the intake form itself applies to a customer's conflicting statements. **Where `recommendation` is `no_bid`**, it is a recommendation, not a decision. Proceed only if the estimator says to, and say so in the quote's assumptions when you do. **Check `bid_decision_state`, and each line's `line_bid_decision`, never context.** `declined` blocks the inquiry, or that line, listed as not quoted by the estimator's decision; reopening is a new decision. `draft`, or a set `bid_decision_reset`, means no ruling covers it: quote only if told to in the run, and record on the quote that it proceeded on an unconfirmed recommendation. **`inquiry_state` is a different axis, and reading it as a decision is the error to avoid.** `superseded` blocks — a later revision exists, quote that one. `stale` does **not** block and never means declined: it means the customer went quiet on open questions. Quote with exclusions naming those questions, or do not quote — but never treat the silence as a no. **Where `recommendation_provisional` is true**, answers were still outstanding when this was scored. Quote if the estimator says to, and carry the outstanding questions onto the assumptions list — they are precisely what the price rests on. Where `hard_disqualifier_stopped` is true, the plant cannot build this as specified: quote only on an explicit override, with the disqualifying constraint on the exclusions list. **Read the geometry estimate's `approval.status` the same way.** `confirmed` means the estimator passed Gate 3 on this model, tolerance reading and operation list. **`draft` is unconfirmed: no hours derived from it reach a quote** unless the estimator says so in this run — then the internal sheet records who said so and that the model was unconfirmed. A `drawing_revision_priced_to` that differs from the intake line's `revision` blocks: that estimate is another part. ### Step 2 — Find comparables Prior jobs for the same part number first, then similar parts by material, process and size. **Show what made each one comparable and how the new job differs.** Same part, half the quantity, different finish, eighteen months of material inflation — the estimator needs those differences to weigh the comparable at all. Where no comparable exists, say so clearly and **do not produce a confident number off nothing.** **Count the history in code.** How many orders, how many for this customer, how many closed work orders, across what dates — each is a filter run over the data, and each goes in the figures block (`../../shared/computed-figures.md`) with that filter. "Consistent across five orders" written from memory misses the sixth, and the one missed tends to be the customer's own. ### Step 3 — Flag capability, then derive the hours **Capability flags come before any price.** Carry the intake form's `constraint_violations`, and check every tolerance on the geometry estimate against demonstrated capability in the Plant context: **a tolerance tighter than the plant has demonstrated on that process is flagged by name, value and capability**, on the internal sheet and, as a plain statement of what is quoted, in the assumptions. Pricing it first and flagging it after is how a plant commits to a tolerance it has never held. **Then hours per operation, from the ladder.** Read `reference/hours-ladder.md` before deriving any hour. Per operation, best available first: 1. **A routing for this part and revision**, cited. A routing for another revision is superseded — shown as such, never priced from. 2. **Closed-job actuals** not flagged `probable_standard`. 3. **Geometry-derived** from the confirmed estimate: the CAM path where Fusion and plant cutting data exist, the feature path always. 4. **First principles**, naming what would firm it up. **`probable_standard` actuals never outrank rung 3.** **Library feeds (`fusion_default`) are rung 4, labelled "library defaults, not plant data."** Setup and run are carried separately at every rung. Where two rungs exist the variance is computed in code and flagged above the Plant context threshold — or 15%, printed word for word as *"default threshold, not plant-set"* — with its drivers named. Every hour, range and variance goes to the figures block. ### Step 4 — Build up the cost Material with its scrap allowance, labor, machine and burden, setup, outside processing, tooling. **Each line cites its source** — standard cost, an actual purchase price with its PO and date, or the actual cost from a specific prior work order. Labor, machine and setup use the hours from Step 3, rung by rung, never a single hours rule for the whole job. Read `reference/cost-buildup.md`. Where a required rate is not supplied, build to that point and **name exactly which rate is missing and which lines it blocks.** Never substitute a plausible one. **No job-quantity total and no unit price for this job may appear while a required rate is unset — not from standard cost, not from price history, not as a ballpark.** Every route to a number is closed, including the ones that look like something else: - *Standard cost × quantity* is a price by another name. Standard cost may be cited as a material basis per unit, labelled standard; it is never extended by the job quantity into a total while labor or burden is blocked. - *A prior sell price × quantity* is a price too. Prior sales may be listed as history — order, date, quantity, price, one line each — and are never multiplied by the requested quantity, never headed "unit price," never offered as a guide to this job. - *"Just a ballpark"* and *"use a typical shop rate"* get the same answer. Decline the typical rate without naming one, and do not then produce the ballpark some other way. **A placeholder is a rate, and prices the job.** Where the `## Plant context` block carries a rate as a marked placeholder with its assumption — recorded by the plant at onboarding, not invented here — build on it. Label every line it touches, and the cost sheet as a whole, **"placeholder rate — not confirmed"**, and put the placeholder on the assumptions list by name. The gaps list names it, never its value. This state, "quote-build is running on a placeholder," is expected. What is still refused is any rate the model supplies itself, from a benchmark or a lookup. **A confirmed rate prices the same way, without the label.** A rate the rates file records as confirmed — given by a named person on a date and approved at onboarding — is the plant's number, as much as one set in `MFG_*`. Cite it on the internal sheet as *plant profile, confirmed by <name> on <date>*, and leave it off the assumptions and gaps lists. **Still produce everything the rates do not touch:** material at the price actually paid on a named PO with its date; the setup amortization structure at each quantity, with the dollar values blocked; outside processing and tooling where sourced; and the date check against shop load. Name each missing rate and both ways to set it — *"update our labour rate"* to record it on the plant's rates page, or `MFG_LABOR_RATE` / `MFG_BURDEN_RATE` in the environment — say which lines each blocks, and say the quote completes as soon as they are set. ### Step 5 — Apply quantity breaks **Recompute setup amortization at each quantity the customer asked for.** Setup does not scale — it is the same cost across more units, and that is the entire reason unit price drops with volume. Show the setup contribution per unit at each break so the spread is explicable. **Setup and run stay separate at every break, at every rung.** Where a source gives only a lumped total, the split is unknown: say so, and **never re-amortize it** — use the highest rung that carries a split for the break arithmetic. ### Step 6 — Apply margin Per the configured rules, **shown as its own line and never buried.** An estimator negotiating needs to see exactly how much room they have. ### Step 7 — Check the promised date against shop load Read the planning export and answer whether the requested date is achievable given what is already committed, plus material and outside-processing lead time. **The work content is setup plus run hours by work center from each operation's winning rung, and the check names that basis.** **If it cannot be met, say so and give the earliest date that can.** Quoting a date the shop cannot hit is how a plant loses a customer twice. Without a planning export, the quote carries a stated lead time with an explicit **"not checked against shop load"** caveat — never a silent omission. Read `reference/date-and-load.md`. ### Step 8 — Assemble both documents The **customer-facing quote** in the plant's template, and the **internal cost-basis sheet** the estimator keeps and defends. **Emit the customer quote between delimiters, and keep everything internal outside them:** ``` === CUSTOMER QUOTE — SENDABLE === …the quote, and only the quote… === END === ``` The cost-basis sheet is its own delimited document, so both are explicit and neither can be mistaken for the other: ``` === INTERNAL COST SHEET — NOT FOR CUSTOMER === …the buildup, every source, each rate against its source, the hours per operation with rung, path, cutting data basis, setup and run, range, variance and drivers, the comparables, the margin line, the figures block… === END === ``` Notes to the estimator that belong to neither document go outside both blocks. Narrative cost prose with no delimiter is not a cost sheet. **An assumption that reveals pricing policy is internal**: "no customer was named, so the default margin rule applies; a different rule is configured for another customer" is useful to the estimator and must never sit inside the quote. On the internal sheet, show each rate's value against its source — `MFG_LABOR_RATE`, *plant profile, confirmed by <name> on <date>*, or *plant profile, placeholder* — that is what makes the basis defensible. **The customer quote states what was quoted, not how it was estimated.** Its assumptions name the **drawing revision priced to** and the **tolerances as read** — *"priced to PN-4471-B Rev 2; unspecified tolerances ±0.005 in per the drawing's general note"* — and any capability flag as a statement of scope. **Then read the block alone, and reject it if it holds** any hours figure — setup, run, cycle or machine time, per piece or per job — any low, expected or high range, any rung, path or source of an hour (routing, actuals, CAM, feature path, geometry estimate, first principles), any cutting-data basis label (`plant_library`, `estimator_entered`, `fusion_default`, library defaults), any variance, `probable_standard`, or anything from `cost_down_proposals`; any rate value — from an `MFG_*` variable or the plant profile, or a figure that is one times something — the word "margin" **in any form — "margin applied," "standard margin," "at our usual margin," a margin percentage, or a line item called margin** — any description of which pricing rule was used, or any customer's name but the addressee's. The customer sees prices; how they were arrived at is the internal sheet's business. A rejected block is rewritten before anything is sent, not annotated. The full list is `../../shared/secret-config.md`, *The outbound document check*. ### Step 9 — Handle the negotiation loop When a customer counters, **revise the existing quote in place and preserve a change record** of what moved and why. A counter is not a new quote. Treating it as one is how a plant ends up with four versions in circulation and no record of what was conceded. Re-derive the affected lines rather than patching the total, and where a price dropped with no scope change, **say that margin moved.** Re-derive from the stored operation record — **a price counter never reopens CAD.** A new drawing revision is not a counter: it goes back to `rfq-to-quote` step 7. Read `reference/revisions.md`. --- ## Approval gates **Nothing reaches a customer without the estimator releasing it.** This is Chain 1's **fourth** gate, at step 11. Gate 1 is the estimator sending the open questions at step 4; Gate 2 is confirming the bid/no-bid at step 5; Gate 3 is confirming the design, tolerance reading and operation list at step 8 — no geometry-derived hour is priced without it or a recorded override. The quote is drafted, shown with its cost basis, and sent by a person. **Every revision goes through the same gate.** A second version of an approved quote is not pre-approved. **Rates are resolved per `../../shared/secret-config.md` and shown only where the audience rule allows.** Nothing is written back to the ERP; the quote is a document the estimator files and sends. The margin line appears on the internal cost-basis sheet, never on the customer-facing quote. --- ## Output **Every document this skill produces is printed in the reply, in full.** Saving a copy to a file as well is fine, as Word or Excel per `../../shared/replies.md`, never `.txt`; a file *instead of* the reply is not. The person reads the reply, and every delimited block and outbound check lives in it — a document only on disk has been neither seen nor checked. **A record handed to another skill is saved and summarised, never pasted as raw YAML**; memory blocks are written where the plant's memory lives; and the name on an approval is asked, never taken from the account. All three per `../../shared/replies.md`. **The customer-facing quote** — line items, prices, quantity breaks, lead time, validity period, terms (payment, freight, packaging, liability — as the customer stated them, or the plant's standard where one was never stated), and the assumptions and exclusions list. Per `../../shared/artifact-style.md`, a quote is a controlled document and uses the plant's own template captured during `onboarding`, **never the house style.** **The internal cost-basis sheet** — the full buildup with every source cited, the comparables used and how they differ, the margin applied, and every assumption made. The estimator hands this to their manager and defends the price without re-deriving it. **It carries the figures block** — every count of orders, lines and comparables, and every extended figure, with the filter or arithmetic behind it. Internal only; it never appears in the customer quote. **The hours-basis table, on the internal sheet** — per operation: rung, path, cutting-data basis, setup and run with their range, every other rung with its variance, the variance drivers, and the toolpath check where the CAM path ran. Any `cost_down_proposals` from the geometry estimate sit here, internal only. **The operation record** — kept with the quote, so a counter re-prices without reopening anything and the estimate can one day be compared with the job's actuals. Its fields are in `reference/hours-ladder.md`. **The assumptions and exclusions list** — explicit, including any commercial term quoted against the plant's standard because the customer never stated one, or flagged because a stated one clashed with the plant's. This is what protects the plant when the customer's expectations diverge later, and it is the section everyone skips. **Revision history** — when revising, with what changed and why. **The Sales command-center page** — Chain 1's closing step. Commercial performance for the period, carrying the intake record and the quote by reference so the page reads from what was actually sent rather than a separate summary of it. Produced as an artifact you place; nothing is written to a system. The customer-facing version carries no margin line. Read `../../shared/command-center.md`. --- ## Edge cases and failure modes - **No comparable exists** — a genuinely new part. Say so; do not produce a confident number off nothing. - **Material price is volatile** — cite the price as of its date and add a validity caveat to the quote. - **The requested date is impossible** — never quietly quote a date the shop cannot hit. - **A quantity far outside anything the plant has run** — flag the extrapolation explicitly. A price at 50,000 derived from experience at 500 is not a quote. - **Currency or unit mismatches between the RFQ and our cost data** — surface them rather than converting silently. - **A required rate was not supplied** — name which one and which lines it blocks. Never invent it, and produce no total and no unit price for the job by any other route — see Step 4. - **The cost history is stale** — cite each figure as of the date its work order closed. A comparable priced in a different material market is still useful, but only if its age is visible. - **A spec on the intake form has no source reference** — block that spec. An unsourced tolerance is an inference, and inferences get priced and then committed to contractually. - **A commercial term (payment, freight, packaging, liability) was never stated** — quote against the plant's standard for that term and name the assumption on the exclusions list. Never fill it in silently. - **The only routing is for an earlier revision** — superseded. Show it as such; derive from the next rung. Never price the new revision on it. - **Actuals agree to within 1%** — `probable_standard`. Shown, never above a geometry-derived estimate. - **The geometry estimate is still `draft`** — no hours from it reach a quote without the estimator's say-so in the run, recorded on the internal sheet. - **A Plant context value is tagged `assumed`** — use it, label it, and name it as a variance driver where it explains more of a gap than the threshold. --- ## Degradation Runs on the intake form plus pasted cost figures with no connector at all. | Not available | What happens | |---|---| | The ERP (NetSuite or Infor M3) | No standard cost, actual cost or purchase-price history. Ask for material prices and rates directly, and label the whole buildup as supplied rather than sourced | | Closed-work-order cost | Build up from standard cost, labelled **"standard cost (3rd preference, no actual-cost comparable)"** — standard cost does not say whether the last job made money | | The intake form | Treat the request as a minimal intake: part and quantity as the user stated them, every other field null and on the missing register. Build from the best basis available and label it. A recorded placeholder rate counts as a set rate | | Purchase-order history | Price material from a current supplier quote or a supplied figure, with its date | | Planning export | Quote a stated lead time with the explicit "not checked against shop load" caveat | | Prior quotes | No precedent for this customer. State it rather than reading absence as a new relationship | | The plant's quote template | Use a neutral quote structure covering every section above, and flag that the controlled template was unavailable | | Routings and actuals | The ladder starts at rung 3, or rung 4 where there is no confirmed geometry estimate. Name both tabs missing | | Geometry estimate | No rung 3 for that line. Say the design step did not run and why; the ladder runs on what exists | | A drawing (path A, text specs only) | **A ROM, never a firm quote**: ROM on the internal sheet, a budgetary estimate on the customer document, nothing about method. An override does not make it firm. `reference/hours-ladder.md`, *Path A* | | Fusion, or its Manufacturing workspace | Feature path only, and the internal sheet says so. No toolpath check is claimed | | Plant cutting data | The CAM path runs on library feeds, labelled `fusion_default`, rung 4. The feature path cannot time a cut; that op is rung 4 | | Machining data in the Plant context | No plant setup or tool-change values, no cell map, and a variance threshold of 15% printed as "default threshold, not plant-set" | --- ## Chain position **Chain 1, RFQ to quote** — the final step, depending on `customer-inquiry-intake` and `data-consolidate` as two independent inputs, not a pipe from one through the other. Both this step and the intake's bid/no-bid depend on cost history. **If the ERP is not connected, both degrade to document-based comparables and the chain says so** rather than quietly producing weaker output. **Metric.** Quote turnaround; hit rate; the unbid book. --- ## Reference files | File | Load when | |---|---| | `reference/hours-ladder.md` | Step 3 — always, before any hour: the ladder, setup and run, ranges, variance, the CAM and feature paths, the per-op sheet columns and the operation record | | `reference/cost-buildup.md` | Steps 2 to 6 — the seven lines, their sources, comparables, and quantity breaks | | `reference/date-and-load.md` | Step 7 — the load check, its scope boundary, and the caveat when it cannot run | | `reference/revisions.md` | Step 9 — the change record, re-deriving from the operation record, and why a new drawing revision is not a counter | | `../../shared/contracts/geometry-estimate.md` | Steps 1 and 3 — the geometry estimate, its approval state and what blocks it | | `../../shared/computed-figures.md` | Steps 2 and 7 — the figures block every stated count is quoted from | | `../../shared/contracts/intake-form.md` | Step 1 — the intake shape, a line with no quantity or no material, the missing behavior for a null commercial term, an ambiguous register entry, uncaptured quantity breaks, a `no_bid` recommendation, the `bid_decision_state` gate, and absent `field_sources` |
© anthropics · Apache-2.0 · anthropics/claude-plugins-official