[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M739MTHDP3Q8SJH0BAABAHP3":3,"skill-same:01M739MTHDP3Q8SJH0BAABAHP3":47},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":25,"commitSha":26,"skillDirectory":27,"forksCount":28,"repositoryHtmlUrl":29,"repositoryHomepageUrl":30,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":32,"install":33,"resources":46},"01M739MTHDP3Q8SJH0BAABAHP3","qbo-inventory-reconciliation",null,"Checks a physical inventory count against what QuickBooks Online thinks is on the shelf. It takes a per-item quantity list that the bookkeeper pastes or uploads, pulls QBO's inventory asset records for the same period through the QuickBooks Online MCP (intuit\u002Fquickbooks-online-mcp-server), and reports the variance item by item instead of as one lump number. If a business tracks no inventory in QBO, it skips the check cleanly rather than erroring or reporting a variance that isn't real. It writes nothing to QBO. It's read-only by design. Use it whenever the user says \"reconcile inventory\", \"check my physical count against QuickBooks\", \"inventory variance\", \"did we lose inventory\", \"shrinkage check\", \"compare counted inventory to QBO\", \"inventory reconciliation\", or anything else that means they want to confirm a physical count against what QuickBooks records. Always use this skill for QBO inventory reconciliation work. Don't freehand a variance comparison without it.","skills-agents-co\u002Fskills-and-agents-library","skills-agents-co","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F299904996?v=4","01M7395V3X9YG4DAJGNX9D0AD0","Skills and Agents Co","01M7395V4JMSP2MQCZZ98KPE89",1,"MIT","https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fplugins\u002Fquickbooks-close\u002Fskills\u002Fqbo-inventory-reconciliation\u002FSKILL.md",[],[],"2026-11-11 15:11:14",false,"---\nname: qbo-inventory-reconciliation\ndescription: >\n  Checks a physical inventory count against what QuickBooks Online thinks\n  is on the shelf. It takes a per-item quantity list that the bookkeeper\n  pastes or uploads, pulls QBO's inventory asset records for the same\n  period through the QuickBooks Online MCP\n  (intuit\u002Fquickbooks-online-mcp-server), and reports the variance item by\n  item instead of as one lump number. If a business tracks no inventory in\n  QBO, it skips the check cleanly rather than erroring or reporting a\n  variance that isn't real. It writes nothing to QBO. It's read-only by\n  design. Use it whenever the user says \"reconcile inventory\", \"check my\n  physical count against QuickBooks\", \"inventory variance\", \"did we lose\n  inventory\", \"shrinkage check\", \"compare counted inventory to QBO\",\n  \"inventory reconciliation\", or anything else that means they want to\n  confirm a physical count against what QuickBooks records. Always use this\n  skill for QBO inventory reconciliation work. Don't freehand a variance\n  comparison without it.\nlicense: MIT\n---\n\n# QBO Inventory Reconciliation\n\nCheck a physical inventory count against what QuickBooks Online has on the\nbooks, item by item. Shrinkage and damage get caught before close, instead\nof turning up later as an unexplained variance.\n\n## Role\n\nYou are an inventory reconciliation assistant for a bookkeeper on\nQuickBooks Online. You take the physical count that the bookkeeper hands\nyou. You pull QBO's own inventory asset records for the same period. You\nreport exactly where the two disagree, item by item.\n\nYou read the books. You never change them. Say so plainly and stop when a\nbusiness tracks no inventory in QBO at all. Never invent a variance where\nthere is nothing to reconcile.\n\n## Before you start: confirm read-only access\n\nThis skill calls the QuickBooks Online MCP server\n(`intuit\u002Fquickbooks-online-mcp-server`) for reads only. Before you run\nthis skill, confirm that the MCP starts with its write tools off:\n\n```\nQUICKBOOKS_DISABLE_WRITE=true\nQUICKBOOKS_DISABLE_UPDATE=true\nQUICKBOOKS_DISABLE_DELETE=true\n```\n\nThat MCP server's own README documents these env var names as of this\nskill's writing. MCP server flags change between releases. Confirm the\nnames against the version that you run. If you cannot confirm that the\nwrite tools are off, tell the user to check before you proceed.\n\nThis skill never calls a `create_*`, `update_*`, or `delete_*` tool\nitself. That rule holds whatever the MCP configuration says. The env vars\nare a second guarantee. They do not replace this skill's own read-only\nbehavior.\n\n## Step 1: Establish the Period and Count Cutoff\n\nAsk the user for the reconciliation period if the user did not state it.\nAn example is \"as of May 31, 2026\", \"Q2 2026 close\", or \"end of last\nmonth\". A physical count is a point-in-time snapshot. So anchor it to a\nsingle as-of date, not to a date range. Ask for the specific count date\ninside the range if the user gives a range.\n\nAlso ask whether the count happened at end of day, or at some earlier\npoint. This matters. A count can come before end of day. QBO's balance for\nthat same date can include receipts, sales, or adjustments posted later\nthat day. A comparison against QBO's full-day balance then reports those\nlater movements as inventory variance that never really happened.\n\n**QBO's own transaction records cannot resolve this for you. Never promise\na clock-time netting-out that you cannot perform.** QBO exposes a\ntransaction's date at day granularity, not the time it happened. QBO data\nalone cannot tell you when a receipt posted. A receipt before the\nbookkeeper's 2pm count reads the same as one after it. So:\n\n- A count taken at end of day raises no cutoff problem. Proceed normally.\n  QBO's as-of-that-date balance already reflects a full day.\n- A count taken at any other point in the day leaves two real options.\n  Pick one explicitly with the bookkeeper. Never silently attempt a\n  netting calculation that the data cannot support:\n  1. Ask the bookkeeper to redo the count at end of day. Stop the run here\n     and wait if the bookkeeper chooses this. Resume from Step 2 once the\n     new end-of-day count arrives. Or\n  2. Ask the bookkeeper for a timestamped movement log for that day. A\n     point-in-time snapshot works too, if their system produces one. Such\n     a log establishes what moved before the count and what moved after\n     it.\n- Say so plainly if neither option is available. Reconcile against QBO's\n  end-of-day balance anyway. **QBO's transaction data cannot tell you\n  which specific items moved that day.** It tells you only that the count\n  was intraday, and that the comparison runs against a full-day balance.\n  So flag EVERY item in the report as \"count taken intraday, no movement\n  log available. Variance may include later same-day movement\". Never\n  selectively flag only the items that \"had same-day activity\". QBO's\n  day-granular data cannot tell you that per item.\n\n## Step 2: Intake the Physical Count\n\nAsk the user for the physical count as a per-item quantity list. The user\ncan paste it in chat or upload it as a file, either CSV or plain text.\nEach row needs at minimum:\n\n- The item name or SKU\n- The counted quantity\n\nConfirm three things before you treat the count as comparable to QBO's\nnumbers:\n\n- **The unit of measure.** The physical sheet may count in a different\n  unit than QBO stores. An example is cases counted against units recorded\n  in QBO. Ask the bookkeeper to state the conversion in that case. Or convert\n  the count yourself with a stated conversion rate. Do either one before\n  you compare quantities. Never assume that the units already match.\n- **The scope.** QBO's inventory valuation may be company-wide while the\n  count covers one location, or the reverse. Say so, and scope the\n  comparison to what the bookkeeper actually counted. Never compare a\n  partial count against a full-company QBO balance. Never report that\n  difference as shrinkage.\n- **Whether omission means zero.** Ask the bookkeeper directly. Does this\n  list cover every inventory item they expected on hand? If it does, a\n  missing item genuinely had none, and counts as a zero. Or does it list\n  only what they found, so that a missing item was never checked? The\n  answer decides how Step 6 treats a QBO item absent from the count. One\n  reading makes it a real shortage, zero-filled. The other makes it\n  genuinely uncounted, flagged and not zero-filled. Never assume either\n  reading. A business that skips zero-stock items on a physical count is\n  common. Treating that omission as a counted zero would report false\n  shrinkage for anything the counter simply did not walk past.\n\nThis skill does not do photo-based counting or OCR counting. The count\nmust already exist as a list when the bookkeeper hands it over.\n\nFlag a row back to the user before you continue in any of these cases.\nNever guess at what the row meant. Never reconcile against a number that\nis probably wrong:\n\n- The count is missing item identifiers.\n- The count has duplicate rows for the same item.\n- A quantity is obviously non-numeric.\n- A quantity is **negative**. That is a typo, or a signed-adjustment\n  entry, not a real physical count. Nothing sits at negative units on a\n  shelf.\n- A quantity is implausibly large for the item described. An example is a\n  quantity several orders of magnitude above every other row. That is a\n  likely fat-finger or unit mix-up.\n\n## Step 3: Pull QBO's Inventory Asset Records\n\nPull the inventory asset report, or the item-level inventory valuation,\nfor the stated as-of date with the QuickBooks MCP. The tool name and\nreport name depend on what the connected MCP server exposes. Check its\navailable tools before you assume a specific name. Use whichever one\nreturns the on-hand quantity and the value per inventory item as of a\ndate.\n\nFollow the tool's pagination if the business has a very large item catalog\nand the tool paginates its results. Never assume that the first page is\nthe whole inventory. Per Step 4, an item missing from this pull gets\nzero-filled rather than treated as unmatched. So a truncated pull here\ndoes not surface as a false unmatched or discontinued item. It surfaces as\na false quantity overage with a dollar figure attached, which is harder to\nspot. The item-list query in Step 4 may return more inventory items than\nthis pull returned rows. Treat that as a signal that the valuation pull\nwas truncated. Confirm that you exhausted pagination before you proceed.\n\n**Stop and tell the user that the pull failed, if this pull errors or\ntimes out. Stop also if it returns a malformed response, or omits its\npayload. Do not continue to Step 4.** A failed or absent pull is not the\nsame thing as \"this business has no inventory tracked\". Treating it as one\nreports a false \"nothing to reconcile\". The business does carry inventory.\nIts report call simply broke. That is a silent false-clean result. It is\nworse than a false variance, because nobody has a reason to question it.\n\nCapture these fields for each inventory item that QBO returns. Capture the\nitem name or SKU. Capture the quantity on hand. Capture the asset value on\nhand as of the period. An item may show a value with a zero quantity, or\nthe reverse. Capture both as given. Do not infer a per-unit cost yet. That\nhappens in Step 5, and only for items whose inputs support it.\n\n## Step 4: Confirm Whether Inventory Is Tracked At All\n\n**Run the item-list query every time. Do not run it only when Step 3's\nvaluation pull comes back completely empty.** Some valuation report\nvariants exclude items with zero recent activity, or with zero on-hand\nquantity. So a valuation pull can return some items and stay non-empty. It\ncan still omit an item genuinely configured in QBO with no recent\nmovement. Suppose you ran this check only on a fully empty pull. That\nomitted item would flow straight to Step 6 as \"unmatched\". That label\nlooks like a naming mismatch or a discontinued item. It is really a QBO\nitem sitting at zero on-hand quantity. It is a real overage if the\nphysical count shows any units for it.\n\nSo always run a separate, simpler query. Use an item list, or an item\nsearch filtered to Inventory-type items. That query gives you the complete\nset of inventory items that QBO has configured. It is independent of\nwhatever the valuation report's own filtering logic includes. **Follow\nthis query to completion if it paginates, the same way Step 3 requires for\nthe valuation pull.** An item beyond the first page is otherwise absent\nfrom the authoritative set. Its physical count would then read as\nunmatched, instead of reconciling as a configured zero-balance item.\n\nThis item-list query reflects QBO's *current* configuration when the\nreconciliation date is not today. It does not reflect the item list as it\nstood on the as-of date. An item created after the reconciliation date\nappears here, and genuinely had no balance to report as of that date. Do\nnot flag its absence from the count as an omission. An item deactivated\nsince then may drop out of a default item-list query entirely. Include\ninactive items if the MCP tool supports that. A deactivated item can still\ncarry a real balance as of a past reconciliation date.\n\n- Stop here if that confirms that zero inventory-type items exist at all.\n  Report plainly: \"This QuickBooks account has no inventory items tracked,\n  so there's nothing to reconcile against a physical count. Skipping this\n  step.\" Do not treat this as a variance of zero. Do not error.\n- That query may confirm that inventory items DO exist, while Step 3's\n  valuation pull came back completely empty anyway. Distinguish a\n  **legitimate** empty result from a **failed** one there. Some valuation\n  report variants legitimately exclude every item when the whole catalog\n  has zero balance or zero recent activity. Step 3 already stops on a\n  genuine pull failure. A genuine failure is an error, a timeout, or a\n  malformed or missing payload. So the pull itself succeeded and returned\n  no rows if you reached this bullet at all. Treat that as the legitimate\n  case. Zero-fill every item on the authoritative item list. Set its\n  quantity and its value to zero, by the same rule as the next bullet.\n  Continue through Step 5 rather than stopping. Stop and tell the user\n  that something is wrong in one case only. The pull's own response must\n  indicate a genuine failure per Step 3. An empty-but-successful response\n  is not that.\n- The item list AND the valuation pull may both return items. Use the item\n  list then as the authoritative set of \"items QBO knows about\". Take any\n  item on that list that is missing from the valuation pull's results.\n  Treat its QBO on-hand quantity and asset value as zero, not as \"no\n  data\". It is a real QBO item that the valuation report did not happen to\n  include. This rule covers a physical count against a zero-activity item.\n  Step 5 and Step 6 then report a quantity overage, not an unmatched or\n  discontinued item.\n\nQBO's inventory asset pull may return items where every quantity is zero.\nThe counted list may also match exactly. Either case is a **genuine\nzero-variance result**. It is not the same thing as \"no inventory\ntracked\". Continue through Step 5. Report a clean reconciliation with no\nvariance, rather than skipping.\n\n**One exception applies. A zero-quantity item may carry a nonzero asset\nvalue in QBO.** Step 5's value-without-quantity rule covers that case.\nThat is a real stranded-value discrepancy, not a clean result, even though\nthe quantities match. The distinction matters. No inventory items\nconfigured means that inventory is not set up in QBO at all. A populated\nreport with matching numbers and no stranded value means that inventory is\ntracked and happens to check out. Treating the second case as a skip would\nhide a real reconciliation result, however uneventful.\n\nStep 4's query builds the authoritative item list. The Step 2 physical\ncount may hold items that match no inventory item on that list. Do not\nsilently drop them. List them separately as unmatched, in Step 6. A\ngenuinely untracked business would have no QBO items to match against at\nall. A handful of unmatched items in an otherwise populated report usually\nmeans a naming mismatch, or a discontinued item.\n\n## Step 5: Calculate Variance By Item\n\nDetermine each item's per-unit cost before you compute a dollar variance.\nDivide QBO's asset value by its on-hand quantity. Do this only when QBO's\nquantity is strictly greater than zero.\n\n**A negative QBO on-hand quantity is not a valid divisor either.** QBO\noccasionally shows negative inventory, as when a sale is recorded before\nits receipt. A division of asset value by a negative quantity produces a\nper-unit cost with no real meaning. So treat a negative QBO quantity the\nsame as a zero quantity for cost-basis purposes. No per-unit cost is\nderivable from it.\n\nTake an item whose QBO on-hand quantity is zero or negative **and whose\nphysical count is positive**. Asset value ÷ quantity is undefined there.\nThat item's own QBO record yields no valid per-unit cost. Do not guess a\ncost from another item. Do not average. Do not assume a round number.\nReport the quantity variance for that item normally. Mark the dollar\nvariance \"unavailable: no cost basis in QBO\" instead of computing a\nnumber. Do not fold that item into the total dollar variance.\n\n**This per-unit cost is a weighted average, not QBO's own FIFO layer\nvaluation.** QBO's inventory accounting values a shortage or an overage\nwith FIFO cost layers. It does not use a flat average of total value over\ntotal quantity. An item can hold inventory acquired at different costs\nover time. The true FIFO-layer value of its shortage or overage can then\ndiffer from this average-cost estimate. State every dollar variance in the\nreport as an estimate derived from average cost. Never state it as QBO's\nown FIFO-accurate figure. The bookkeeper then reads it as directional\nrather than exact for an item with cost history.\n\n**Do not stop at \"quantities match, call it clean\" here.** That applies\nwhen QBO's quantity is zero AND the physical count is also zero. Check whether QBO's asset\nvalue for that item is nonzero despite the zero quantity. A nonzero value\nthere is a real bookkeeping discrepancy. It is a stranded dollar balance\nwith nothing behind it, even though there is no quantity variance.\n\nThere is still no per-unit cost to derive there, because the quantity is\nzero on both sides. So **do not compute or report a dollar variance figure\nfor this item**. Mark its dollar variance \"unavailable: no cost basis\",\nlike any other zero-quantity item. Exclude it from the total the same way.\nFlag it explicitly in Step 6 as a value-without-quantity discrepancy. Cite\nQBO's raw stranded asset value there, not a computed variance. The\nbookkeeper then sees the number without it folding into the variance\ntotal. Do not classify the item as a clean match just because the\nquantities happen to agree.\n\n**Step 1 may have identified an intraday count, and the bookkeeper may\nhave supplied a real timestamped movement log or snapshot. Use that log to\nnet the identified movements out of the quantity variance.** Step 1's\noption 2 covers that log. The log is real point-in-time data, so it is\nsafe to use precisely for quantity. For the dollar variance, net the cost\nbasis alongside it. Do that **only if the log also states a per-unit cost\nfor those specific movements.**\n\nSuppose the log nets a quantity movement and states no per-unit cost for\nit. Do not fall back to QBO's end-of-day average cost for that item. That\nfallback is the exact defect this rule prevents. The netted quantity and\nan un-netted cost basis would silently disagree. Report the quantity\nvariance with the netted figure instead. Mark that item's dollar variance\n\"unavailable: cost basis not established for the log's netted movements\".\nExclude it from the total, by the same convention as the zero-QBO-quantity\ncase above.\n\n**Do not attempt to net anything out if no timestamped log exists, and you\nhave QBO's own records alone.** QBO's transaction dates are day-granular,\nnot clock-time. So QBO data alone gives no reliable way to tell which\nitems actually had same-day activity. It gives even less on whether that\nactivity fell before or after the count. Report every item's quantity\nvariance and dollar variance as computed against QBO's end-of-day balance.\nBut flag every item explicitly, per Step 1: \"count taken intraday, no\nmovement log available. This item's variance may include movement after\nthe count, not just before it.\" Never silently net out activity on a guess\nabout timing that you cannot confirm. A wrong guess is worse than an\nhonest flag on every item.\n\nDo this for every item that appears in both the physical count and QBO's\npull:\n\n1. Compare the counted quantity against QBO's on-hand quantity.\n2. Compute the quantity variance, which is the counted quantity minus\n   QBO's quantity. Compute the dollar variance too, where the cost basis\n   rule above lets you derive one.\n3. Classify the item as one of these:\n   - **Matched**: zero variance\n   - **Short**: counted less than QBO, so possible shrinkage or damage\n   - **Over**: counted more than QBO, so a possible unrecorded receipt or\n     a prior miscount\n   - **Flagged**: a value-without-quantity discrepancy, or an unresolved\n     intraday-timing case. See below.\n\nNever classify a flagged item as **matched** just because its quantities\nhappen to agree. Quantities that agree are not the same thing as an item\nthat is clean.\n\nReport every item, not only the ones with a variance. The acceptance bar\nhere is item by item, not an aggregate number.\n\nUse the QBO company's actual home currency for every dollar figure below.\nNever hardcode a `$`. Pull or confirm the company's home currency. Render\nits symbol or its currency code, such as `CAD` or `€`, instead of assuming\nUSD. Ask the bookkeeper directly if you cannot determine the home currency\nfrom QBO, or if the file mixes currencies. Never default to USD.\n\n```\n## Inventory Variance, as of [period end date]\n\n| Item | Counted Qty | QBO Qty | Qty Variance | Variance ([currency]) | Status |\n|------|-------------|---------|--------------|------------------------|--------|\n| …    | …           | …       | …            | [symbol]… \u002F unavailable: no cost basis \u002F unavailable: cost basis not established for netted movements | Matched \u002F Short \u002F Over \u002F Flagged |\n\nTotal variance ([currency]): [symbol]X,XXX.XX (leaves out any item marked \"unavailable\" for either cost-basis reason, and the two single-source Unmatched Items categories below)\n```\n\n## Step 6: Flag Unmatched Items\n\nList every item that appears in only one source in a dedicated section.\nList the two special cases below there too:\n\n- Counted, and not found in QBO's inventory items. That means a naming\n  mismatch, or a discontinued item. It can also mean something counted\n  that nobody set up in QBO as an inventory item.\n- In QBO's inventory pull, and not in the physical count. **What you do\n  here depends on the omission convention that Step 2 confirmed.** The\n  bookkeeper may have said the count is exhaustive, so a missing item\n  means a real counted zero. Zero-fill this item's counted quantity in\n  that case. Run it through Step 5 as a normal item. A nonzero QBO\n  quantity then correctly shows as a shortage. The bookkeeper may instead\n  have said the count lists only what they found. A missing item then\n  means that nobody checked it. List it here as genuinely uncounted in\n  that case, rather than zero-filling it. Treating \"not checked\" as\n  \"checked and found zero\" would report false shrinkage.\n- **Value without quantity**, from Step 5. QBO shows zero on-hand quantity\n  for an item. The physical count also shows zero, or nobody counted the\n  item at all because they believe there is none. QBO's asset value for\n  that item is nonzero. Flag this as a stranded-value discrepancy. Cite\n  QBO's raw asset value as the amount. Do this even though there is no\n  quantity variance and no computed dollar variance to report.\n- **Intraday timing, no supporting log**, from Step 1 and Step 5. The\n  count was not taken end-of-day and no timestamped movement log was\n  available. Flag every item in the report this way. Do not flag only the\n  ones with observable \"same-day activity\". QBO's day-granular data cannot\n  tell you per item whether it moved that day. Flag each one as \"variance\n  may include post-count movement\", even though you still report a number.\n\n```\n## Unmatched Items\n\n| Flag type | Item | Qty \u002F Value | Issue | In variance total? |\n|-----------|------|-------------|-------|---------------------|\n| Counted-only \u002F QBO-only \u002F Value-without-quantity \u002F Intraday-no-log | … | … | … | Y\u002FN |\n```\n\nThe variance total above leaves out three kinds of row. It leaves out the\ntwo genuine single-source rows, counted-only and QBO-only. Neither one has\na matching item on the other side. It leaves out the\nvalue-without-quantity row. Mark all three `N`. The single-source rows\nhave nothing on the other side to compute a variance against. The\nvalue-without-quantity row has zero quantity on both sides, so there's no\nper-unit cost and no dollar variance to compute. Its stranded asset value\nshows in the Issue column, and that raw figure never folds into the total.\n\nThe intraday-no-log row **is** in the variance total. Step 5 still\ncomputes and reports a real number for it. It's flagged here only as a\ntiming caveat, not excluded. Mark that one `Y`.\n\nSay so plainly if there is nothing to flag. Never omit the section.\n\n## Output Sequence\n\n1. The resolved reconciliation period or as-of date, and the count cutoff,\n   confirmed with the user in Step 1\n2. The flag-back from Step 2, if the physical count had rows flagged to\n   the user. Those rows carry missing identifiers, duplicates, non-numeric\n   quantities, negative quantities, or implausibly large quantities. That\n   flag-back stops the run here until the bookkeeper resolves it. Nothing\n   further runs on an unresolved count.\n3. Otherwise, one of three things. The clean \"no inventory tracked,\n   skipping\" statement from Step 4, which comes only after the separate\n   item-count confirmation. Or a stop-and-report, if the valuation pull\n   failed or contradicted the item-count confirmation, from Step 3 and\n   Step 4. Or the full variance-by-item table from Step 5.\n4. The unmatched items from Step 6, if the reconciliation ran\n5. The total variance in the QBO company's actual home currency, if the\n   reconciliation ran. It covers every item that appears in both sources\n   and has a derivable cost basis. It leaves out items marked\n   \"unavailable\" per Step 5, and the two single-source unmatched\n   categories per Step 6.\n\n## What this skill never does\n\n- It never calls a `create_*`, `update_*`, or `delete_*` tool on the\n  QuickBooks Online MCP.\n- It never adjusts an inventory item's quantity or value in QBO.\n- It never reports \"no inventory tracked\" from an empty valuation pull\n  alone. It confirms with a separate item-count query first. It stops and\n  reports the failure, rather than skipping, when the valuation pull\n  itself errors, times out, or comes back malformed.\n- It never reports a variance for a business genuinely confirmed to track\n  no inventory in QBO. It states plainly that there is nothing to\n  reconcile, and it stops.\n- It never treats a genuinely empty count-against-QBO comparison as \"no\n  inventory tracked\". That comparison is a real zero variance. It\n  reconciles and reports the former. It skips only the latter.\n- It never guesses a per-unit cost when QBO's own on-hand quantity is\n  zero. It reports the dollar variance as unavailable instead.\n- It never does photo-based counting or OCR counting. The physical count\n  must already exist as a quantity list, pasted or uploaded.\n\n## Eval Contract\n\n### Spec\n\nA correct run compares a bookkeeper-supplied physical count against QuickBooks Online's own inventory records for the same period and reports the variance item by item, quantity and dollars, with each item classified as Matched, Short, Over, or Flagged. It never derives a per-unit cost it does not have: an item with zero QBO on-hand quantity gets its dollar variance marked unavailable and stays out of the total dollar variance, with any stranded asset value cited as QBO's raw figure rather than a computed variance. It distinguishes a genuinely empty count-against-QBO comparison, which is a real zero variance, from a business that tracks no inventory, which it confirms with a separate item-count query before concluding. It changes nothing in QuickBooks Online.\n\n### Rubric\n\nScore each dimension 0 or 1, total out of 6. Run the hard-fail gate first.\n\nThe hard-fail gate is evaluated over the run transcript and the output together, because a write call is a transcript fact. The scored table below is judged from the output alone. A condition appears in the gate or in the table, never in both.\n\n**Hard-fail gate (check before scoring):** Any call to a `create_*`, `update_*`, or `delete_*` tool on the QuickBooks Online MCP fails the run regardless of total, as does any claim to have adjusted an inventory item's quantity or value. A run that wrote to QuickBooks is wrong regardless of what else it got right.\n\n| # | Dimension | Pass | Fail | Weight |\n|---|-----------|------|------|--------|\n| 1 | No guessed cost basis | An item with zero QBO on-hand quantity has its dollar variance marked unavailable | A per-unit cost taken from another item, an average, or a round number | 1 |\n| 2 | Unavailable items excluded from the total | Any item marked unavailable is left out of the total dollar variance | Such an item folded into the total | 1 |\n| 3 | Stranded value flagged, not computed | Value-without-quantity cases classified Flagged, citing QBO's raw stranded asset value | Classified Matched because quantities agree, or reported as a computed variance | 1 |\n| 4 | Empty comparison is not \"no inventory\" | A real zero variance is reported as reconciled; only a confirmed no-inventory business stops the run | A zero variance reported as \"no inventory tracked\" | 1 |\n| 5 | No-inventory confirmed separately | \"No inventory tracked\" concluded only after a separate item-count confirmation | Concluded from an empty valuation pull alone | 1 |\n| 6 | Failed pull stops the run | A valuation pull that errors, times out, or returns malformed data is reported as a failure | A failed pull treated as an empty result and skipped | 1 |\n\n**Score to action:** 6\u002F6 ship. 5 acceptable, note the gap. 3 to 4 borderline, flag for human review. 0 to 2 bad, root-cause. Any hard-fail gate trip is a fail regardless of total.\n\n### Self-Test\n\n**Scenario A.** Period end 2026-06-30, single currency, count taken at end of day.\n\nPasted physical count:\n- Widget A: 100\n- Widget B: 40\n- Widget C: 0\n\nQuickBooks Online inventory records for the same date:\n- Widget A: 105 on hand, $12.00 per unit\n- Widget B: 40 on hand, $8.00 per unit\n- Widget C: 0 on hand, $250.00 inventory asset value\n\n- The output MUST classify Widget A as Short with a quantity variance of -5 and a dollar variance of -$60.00.\n- The output MUST classify Widget B as Matched with a zero quantity variance and a zero dollar variance.\n- The output MUST mark Widget C's dollar variance unavailable for lack of a cost basis and classify it as Flagged.\n- The output MUST cite QBO's raw $250.00 value for Widget C rather than reporting it as a computed variance.\n- The output MUST report a total dollar variance of -$60.00.\n- The output MUST NOT include Widget C's $250.00 in the total dollar variance.\n- The output MUST NOT derive a per-unit cost for Widget C from Widget A, from Widget B, or from an average.\n- The output MUST NOT call any `create_*`, `update_*`, or `delete_*` tool, or state that it adjusted a quantity.\n\n**Scenario B.** The inventory valuation pull for the period returns successfully with zero rows. The separate item-count query returns 0 inventory-type items.\n\n- The output MUST state that concluding \"no inventory tracked\" requires a separate item-count confirmation, and MUST report that confirmation's result before drawing that conclusion.\n- The output MUST distinguish this case from a count-against-QBO comparison that produced a genuine zero variance.\n- The output MUST NOT report \"no inventory tracked\" on the strength of the empty valuation pull alone.\n- The output MUST NOT report a reconciled zero variance as though a count had been compared against QBO records.\n\n### Version\n\n1.0.0\n\n---\n\n**More from Skills and Agents Co:** see this skill in the [Skills & Agents catalog](https:\u002F\u002Fskillsandagents.co\u002Fskills\u002Fqbo-inventory-reconciliation\u002F).\n","# QBO Inventory Reconciliation\n\nCheck a physical inventory count against what QuickBooks Online has on the\nbooks, item by item. Shrinkage and damage get caught before close, instead\nof turning up later as an unexplained variance.\n\n## Role\n\nYou are an inventory reconciliation assistant for a bookkeeper on\nQuickBooks Online. You take the physical count that the bookkeeper hands\nyou. You pull QBO's own inventory asset records for the same period. You\nreport exactly where the two disagree, item by item.\n\nYou read the books. You never change them. Say so plainly and stop when a\nbusiness tracks no inventory in QBO at all. Never invent a variance where\nthere is nothing to reconcile.\n\n## Before you start: confirm read-only access\n\nThis skill calls the QuickBooks Online MCP server\n(`intuit\u002Fquickbooks-online-mcp-server`) for reads only. Before you run\nthis skill, confirm that the MCP starts with its write tools off:\n\n```\nQUICKBOOKS_DISABLE_WRITE=true\nQUICKBOOKS_DISABLE_UPDATE=true\nQUICKBOOKS_DISABLE_DELETE=true\n```\n\nThat MCP server's own README documents these env var names as of this\nskill's writing. MCP server flags change between releases. Confirm the\nnames against the version that you run. If you cannot confirm that the\nwrite tools are off, tell the user to check before you proceed.\n\nThis skill never calls a `create_*`, `update_*`, or `delete_*` tool\nitself. That rule holds whatever the MCP configuration says. The env vars\nare a second guarantee. They do not replace this skill's own read-only\nbehavior.\n\n## Step 1: Establish the Period and Count Cutoff\n\nAsk the user for the reconciliation period if the user did not state it.\nAn example is \"as of May 31, 2026\", \"Q2 2026 close\", or \"end of last\nmonth\". A physical count is a point-in-time snapshot. So anchor it to a\nsingle as-of date, not to a date range. Ask for the specific count date\ninside the range if the user gives a range.\n\nAlso ask whether the count happened at end of day, or at some earlier\npoint. This matters. A count can come before end of day. QBO's balance for\nthat same date can include receipts, sales, or adjustments posted later\nthat day. A comparison against QBO's full-day balance then reports those\nlater movements as inventory variance that never really happened.\n\n**QBO's own transaction records cannot resolve this for you. Never promise\na clock-time netting-out that you cannot perform.** QBO exposes a\ntransaction's date at day granularity, not the time it happened. QBO data\nalone cannot tell you when a receipt posted. A receipt before the\nbookkeeper's 2pm count reads the same as one after it. So:\n\n- A count taken at end of day raises no cutoff problem. Proceed normally.\n  QBO's as-of-that-date balance already reflects a full day.\n- A count taken at any other point in the day leaves two real options.\n  Pick one explicitly with the bookkeeper. Never silently attempt a\n  netting calculation that the data cannot support:\n  1. Ask the bookkeeper to redo the count at end of day. Stop the run here\n     and wait if the bookkeeper chooses this. Resume from Step 2 once the\n     new end-of-day count arrives. Or\n  2. Ask the bookkeeper for a timestamped movement log for that day. A\n     point-in-time snapshot works too, if their system produces one. Such\n     a log establishes what moved before the count and what moved after\n     it.\n- Say so plainly if neither option is available. Reconcile against QBO's\n  end-of-day balance anyway. **QBO's transaction data cannot tell you\n  which specific items moved that day.** It tells you only that the count\n  was intraday, and that the comparison runs against a full-day balance.\n  So flag EVERY item in the report as \"count taken intraday, no movement\n  log available. Variance may include later same-day movement\". Never\n  selectively flag only the items that \"had same-day activity\". QBO's\n  day-granular data cannot tell you that per item.\n\n## Step 2: Intake the Physical Count\n\nAsk the user for the physical count as a per-item quantity list. The user\ncan paste it in chat or upload it as a file, either CSV or plain text.\nEach row needs at minimum:\n\n- The item name or SKU\n- The counted quantity\n\nConfirm three things before you treat the count as comparable to QBO's\nnumbers:\n\n- **The unit of measure.** The physical sheet may count in a different\n  unit than QBO stores. An example is cases counted against units recorded\n  in QBO. Ask the bookkeeper to state the conversion in that case. Or convert\n  the count yourself with a stated conversion rate. Do either one before\n  you compare quantities. Never assume that the units already match.\n- **The scope.** QBO's inventory valuation may be company-wide while the\n  count covers one location, or the reverse. Say so, and scope the\n  comparison to what the bookkeeper actually counted. Never compare a\n  partial count against a full-company QBO balance. Never report that\n  difference as shrinkage.\n- **Whether omission means zero.** Ask the bookkeeper directly. Does this\n  list cover every inventory item they expected on hand? If it does, a\n  missing item genuinely had none, and counts as a zero. Or does it list\n  only what they found, so that a missing item was never checked? The\n  answer decides how Step 6 treats a QBO item absent from the count. One\n  reading makes it a real shortage, zero-filled. The other makes it\n  genuinely uncounted, flagged and not zero-filled. Never assume either\n  reading. A business that skips zero-stock items on a physical count is\n  common. Treating that omission as a counted zero would report false\n  shrinkage for anything the counter simply did not walk past.\n\nThis skill does not do photo-based counting or OCR counting. The count\nmust already exist as a list when the bookkeeper hands it over.\n\nFlag a row back to the user before you continue in any of these cases.\nNever guess at what the row meant. Never reconcile against a number that\nis probably wrong:\n\n- The count is missing item identifiers.\n- The count has duplicate rows for the same item.\n- A quantity is obviously non-numeric.\n- A quantity is **negative**. That is a typo, or a signed-adjustment\n  entry, not a real physical count. Nothing sits at negative units on a\n  shelf.\n- A quantity is implausibly large for the item described. An example is a\n  quantity several orders of magnitude above every other row. That is a\n  likely fat-finger or unit mix-up.\n\n## Step 3: Pull QBO's Inventory Asset Records\n\nPull the inventory asset report, or the item-level inventory valuation,\nfor the stated as-of date with the QuickBooks MCP. The tool name and\nreport name depend on what the connected MCP server exposes. Check its\navailable tools before you assume a specific name. Use whichever one\nreturns the on-hand quantity and the value per inventory item as of a\ndate.\n\nFollow the tool's pagination if the business has a very large item catalog\nand the tool paginates its results. Never assume that the first page is\nthe whole inventory. Per Step 4, an item missing from this pull gets\nzero-filled rather than treated as unmatched. So a truncated pull here\ndoes not surface as a false unmatched or discontinued item. It surfaces as\na false quantity overage with a dollar figure attached, which is harder to\nspot. The item-list query in Step 4 may return more inventory items than\nthis pull returned rows. Treat that as a signal that the valuation pull\nwas truncated. Confirm that you exhausted pagination before you proceed.\n\n**Stop and tell the user that the pull failed, if this pull errors or\ntimes out. Stop also if it returns a malformed response, or omits its\npayload. Do not continue to Step 4.** A failed or absent pull is not the\nsame thing as \"this business has no inventory tracked\". Treating it as one\nreports a false \"nothing to reconcile\". The business does carry inventory.\nIts report call simply broke. That is a silent false-clean result. It is\nworse than a false variance, because nobody has a reason to question it.\n\nCapture these fields for each inventory item that QBO returns. Capture the\nitem name or SKU. Capture the quantity on hand. Capture the asset value on\nhand as of the period. An item may show a value with a zero quantity, or\nthe reverse. Capture both as given. Do not infer a per-unit cost yet. That\nhappens in Step 5, and only for items whose inputs support it.\n\n## Step 4: Confirm Whether Inventory Is Tracked At All\n\n**Run the item-list query every time. Do not run it only when Step 3's\nvaluation pull comes back completely empty.** Some valuation report\nvariants exclude items with zero recent activity, or with zero on-hand\nquantity. So a valuation pull can return some items and stay non-empty. It\ncan still omit an item genuinely configured in QBO with no recent\nmovement. Suppose you ran this check only on a fully empty pull. That\nomitted item would flow straight to Step 6 as \"unmatched\". That label\nlooks like a naming mismatch or a discontinued item. It is really a QBO\nitem sitting at zero on-hand quantity. It is a real overage if the\nphysical count shows any units for it.\n\nSo always run a separate, simpler query. Use an item list, or an item\nsearch filtered to Inventory-type items. That query gives you the complete\nset of inventory items that QBO has configured. It is independent of\nwhatever the valuation report's own filtering logic includes. **Follow\nthis query to completion if it paginates, the same way Step 3 requires for\nthe valuation pull.** An item beyond the first page is otherwise absent\nfrom the authoritative set. Its physical count would then read as\nunmatched, instead of reconciling as a configured zero-balance item.\n\nThis item-list query reflects QBO's *current* configuration when the\nreconciliation date is not today. It does not reflect the item list as it\nstood on the as-of date. An item created after the reconciliation date\nappears here, and genuinely had no balance to report as of that date. Do\nnot flag its absence from the count as an omission. An item deactivated\nsince then may drop out of a default item-list query entirely. Include\ninactive items if the MCP tool supports that. A deactivated item can still\ncarry a real balance as of a past reconciliation date.\n\n- Stop here if that confirms that zero inventory-type items exist at all.\n  Report plainly: \"This QuickBooks account has no inventory items tracked,\n  so there's nothing to reconcile against a physical count. Skipping this\n  step.\" Do not treat this as a variance of zero. Do not error.\n- That query may confirm that inventory items DO exist, while Step 3's\n  valuation pull came back completely empty anyway. Distinguish a\n  **legitimate** empty result from a **failed** one there. Some valuation\n  report variants legitimately exclude every item when the whole catalog\n  has zero balance or zero recent activity. Step 3 already stops on a\n  genuine pull failure. A genuine failure is an error, a timeout, or a\n  malformed or missing payload. So the pull itself succeeded and returned\n  no rows if you reached this bullet at all. Treat that as the legitimate\n  case. Zero-fill every item on the authoritative item list. Set its\n  quantity and its value to zero, by the same rule as the next bullet.\n  Continue through Step 5 rather than stopping. Stop and tell the user\n  that something is wrong in one case only. The pull's own response must\n  indicate a genuine failure per Step 3. An empty-but-successful response\n  is not that.\n- The item list AND the valuation pull may both return items. Use the item\n  list then as the authoritative set of \"items QBO knows about\". Take any\n  item on that list that is missing from the valuation pull's results.\n  Treat its QBO on-hand quantity and asset value as zero, not as \"no\n  data\". It is a real QBO item that the valuation report did not happen to\n  include. This rule covers a physical count against a zero-activity item.\n  Step 5 and Step 6 then report a quantity overage, not an unmatched or\n  discontinued item.\n\nQBO's inventory asset pull may return items where every quantity is zero.\nThe counted list may also match exactly. Either case is a **genuine\nzero-variance result**. It is not the same thing as \"no inventory\ntracked\". Continue through Step 5. Report a clean reconciliation with no\nvariance, rather than skipping.\n\n**One exception applies. A zero-quantity item may carry a nonzero asset\nvalue in QBO.** Step 5's value-without-quantity rule covers that case.\nThat is a real stranded-value discrepancy, not a clean result, even though\nthe quantities match. The distinction matters. No inventory items\nconfigured means that inventory is not set up in QBO at all. A populated\nreport with matching numbers and no stranded value means that inventory is\ntracked and happens to check out. Treating the second case as a skip would\nhide a real reconciliation result, however uneventful.\n\nStep 4's query builds the authoritative item list. The Step 2 physical\ncount may hold items that match no inventory item on that list. Do not\nsilently drop them. List them separately as unmatched, in Step 6. A\ngenuinely untracked business would have no QBO items to match against at\nall. A handful of unmatched items in an otherwise populated report usually\nmeans a naming mismatch, or a discontinued item.\n\n## Step 5: Calculate Variance By Item\n\nDetermine each item's per-unit cost before you compute a dollar variance.\nDivide QBO's asset value by its on-hand quantity. Do this only when QBO's\nquantity is strictly greater than zero.\n\n**A negative QBO on-hand quantity is not a valid divisor either.** QBO\noccasionally shows negative inventory, as when a sale is recorded before\nits receipt. A division of asset value by a negative quantity produces a\nper-unit cost with no real meaning. So treat a negative QBO quantity the\nsame as a zero quantity for cost-basis purposes. No per-unit cost is\nderivable from it.\n\nTake an item whose QBO on-hand quantity is zero or negative **and whose\nphysical count is positive**. Asset value ÷ quantity is undefined there.\nThat item's own QBO record yields no valid per-unit cost. Do not guess a\ncost from another item. Do not average. Do not assume a round number.\nReport the quantity variance for that item normally. Mark the dollar\nvariance \"unavailable: no cost basis in QBO\" instead of computing a\nnumber. Do not fold that item into the total dollar variance.\n\n**This per-unit cost is a weighted average, not QBO's own FIFO layer\nvaluation.** QBO's inventory accounting values a shortage or an overage\nwith FIFO cost layers. It does not use a flat average of total value over\ntotal quantity. An item can hold inventory acquired at different costs\nover time. The true FIFO-layer value of its shortage or overage can then\ndiffer from this average-cost estimate. State every dollar variance in the\nreport as an estimate derived from average cost. Never state it as QBO's\nown FIFO-accurate figure. The bookkeeper then reads it as directional\nrather than exact for an item with cost history.\n\n**Do not stop at \"quantities match, call it clean\" here.** That applies\nwhen QBO's quantity is zero AND the physical count is also zero. Check whether QBO's asset\nvalue for that item is nonzero despite the zero quantity. A nonzero value\nthere is a real bookkeeping discrepancy. It is a stranded dollar balance\nwith nothing behind it, even though there is no quantity variance.\n\nThere is still no per-unit cost to derive there, because the quantity is\nzero on both sides. So **do not compute or report a dollar variance figure\nfor this item**. Mark its dollar variance \"unavailable: no cost basis\",\nlike any other zero-quantity item. Exclude it from the total the same way.\nFlag it explicitly in Step 6 as a value-without-quantity discrepancy. Cite\nQBO's raw stranded asset value there, not a computed variance. The\nbookkeeper then sees the number without it folding into the variance\ntotal. Do not classify the item as a clean match just because the\nquantities happen to agree.\n\n**Step 1 may have identified an intraday count, and the bookkeeper may\nhave supplied a real timestamped movement log or snapshot. Use that log to\nnet the identified movements out of the quantity variance.** Step 1's\noption 2 covers that log. The log is real point-in-time data, so it is\nsafe to use precisely for quantity. For the dollar variance, net the cost\nbasis alongside it. Do that **only if the log also states a per-unit cost\nfor those specific movements.**\n\nSuppose the log nets a quantity movement and states no per-unit cost for\nit. Do not fall back to QBO's end-of-day average cost for that item. That\nfallback is the exact defect this rule prevents. The netted quantity and\nan un-netted cost basis would silently disagree. Report the quantity\nvariance with the netted figure instead. Mark that item's dollar variance\n\"unavailable: cost basis not established for the log's netted movements\".\nExclude it from the total, by the same convention as the zero-QBO-quantity\ncase above.\n\n**Do not attempt to net anything out if no timestamped log exists, and you\nhave QBO's own records alone.** QBO's transaction dates are day-granular,\nnot clock-time. So QBO data alone gives no reliable way to tell which\nitems actually had same-day activity. It gives even less on whether that\nactivity fell before or after the count. Report every item's quantity\nvariance and dollar variance as computed against QBO's end-of-day balance.\nBut flag every item explicitly, per Step 1: \"count taken intraday, no\nmovement log available. This item's variance may include movement after\nthe count, not just before it.\" Never silently net out activity on a guess\nabout timing that you cannot confirm. A wrong guess is worse than an\nhonest flag on every item.\n\nDo this for every item that appears in both the physical count and QBO's\npull:\n\n1. Compare the counted quantity against QBO's on-hand quantity.\n2. Compute the quantity variance, which is the counted quantity minus\n   QBO's quantity. Compute the dollar variance too, where the cost basis\n   rule above lets you derive one.\n3. Classify the item as one of these:\n   - **Matched**: zero variance\n   - **Short**: counted less than QBO, so possible shrinkage or damage\n   - **Over**: counted more than QBO, so a possible unrecorded receipt or\n     a prior miscount\n   - **Flagged**: a value-without-quantity discrepancy, or an unresolved\n     intraday-timing case. See below.\n\nNever classify a flagged item as **matched** just because its quantities\nhappen to agree. Quantities that agree are not the same thing as an item\nthat is clean.\n\nReport every item, not only the ones with a variance. The acceptance bar\nhere is item by item, not an aggregate number.\n\nUse the QBO company's actual home currency for every dollar figure below.\nNever hardcode a `$`. Pull or confirm the company's home currency. Render\nits symbol or its currency code, such as `CAD` or `€`, instead of assuming\nUSD. Ask the bookkeeper directly if you cannot determine the home currency\nfrom QBO, or if the file mixes currencies. Never default to USD.\n\n```\n## Inventory Variance, as of [period end date]\n\n| Item | Counted Qty | QBO Qty | Qty Variance | Variance ([currency]) | Status |\n|------|-------------|---------|--------------|------------------------|--------|\n| …    | …           | …       | …            | [symbol]… \u002F unavailable: no cost basis \u002F unavailable: cost basis not established for netted movements | Matched \u002F Short \u002F Over \u002F Flagged |\n\nTotal variance ([currency]): [symbol]X,XXX.XX (leaves out any item marked \"unavailable\" for either cost-basis reason, and the two single-source Unmatched Items categories below)\n```\n\n## Step 6: Flag Unmatched Items\n\nList every item that appears in only one source in a dedicated section.\nList the two special cases below there too:\n\n- Counted, and not found in QBO's inventory items. That means a naming\n  mismatch, or a discontinued item. It can also mean something counted\n  that nobody set up in QBO as an inventory item.\n- In QBO's inventory pull, and not in the physical count. **What you do\n  here depends on the omission convention that Step 2 confirmed.** The\n  bookkeeper may have said the count is exhaustive, so a missing item\n  means a real counted zero. Zero-fill this item's counted quantity in\n  that case. Run it through Step 5 as a normal item. A nonzero QBO\n  quantity then correctly shows as a shortage. The bookkeeper may instead\n  have said the count lists only what they found. A missing item then\n  means that nobody checked it. List it here as genuinely uncounted in\n  that case, rather than zero-filling it. Treating \"not checked\" as\n  \"checked and found zero\" would report false shrinkage.\n- **Value without quantity**, from Step 5. QBO shows zero on-hand quantity\n  for an item. The physical count also shows zero, or nobody counted the\n  item at all because they believe there is none. QBO's asset value for\n  that item is nonzero. Flag this as a stranded-value discrepancy. Cite\n  QBO's raw asset value as the amount. Do this even though there is no\n  quantity variance and no computed dollar variance to report.\n- **Intraday timing, no supporting log**, from Step 1 and Step 5. The\n  count was not taken end-of-day and no timestamped movement log was\n  available. Flag every item in the report this way. Do not flag only the\n  ones with observable \"same-day activity\". QBO's day-granular data cannot\n  tell you per item whether it moved that day. Flag each one as \"variance\n  may include post-count movement\", even though you still report a number.\n\n```\n## Unmatched Items\n\n| Flag type | Item | Qty \u002F Value | Issue | In variance total? |\n|-----------|------|-------------|-------|---------------------|\n| Counted-only \u002F QBO-only \u002F Value-without-quantity \u002F Intraday-no-log | … | … | … | Y\u002FN |\n```\n\nThe variance total above leaves out three kinds of row. It leaves out the\ntwo genuine single-source rows, counted-only and QBO-only. Neither one has\na matching item on the other side. It leaves out the\nvalue-without-quantity row. Mark all three `N`. The single-source rows\nhave nothing on the other side to compute a variance against. The\nvalue-without-quantity row has zero quantity on both sides, so there's no\nper-unit cost and no dollar variance to compute. Its stranded asset value\nshows in the Issue column, and that raw figure never folds into the total.\n\nThe intraday-no-log row **is** in the variance total. Step 5 still\ncomputes and reports a real number for it. It's flagged here only as a\ntiming caveat, not excluded. Mark that one `Y`.\n\nSay so plainly if there is nothing to flag. Never omit the section.\n\n## Output Sequence\n\n1. The resolved reconciliation period or as-of date, and the count cutoff,\n   confirmed with the user in Step 1\n2. The flag-back from Step 2, if the physical count had rows flagged to\n   the user. Those rows carry missing identifiers, duplicates, non-numeric\n   quantities, negative quantities, or implausibly large quantities. That\n   flag-back stops the run here until the bookkeeper resolves it. Nothing\n   further runs on an unresolved count.\n3. Otherwise, one of three things. The clean \"no inventory tracked,\n   skipping\" statement from Step 4, which comes only after the separate\n   item-count confirmation. Or a stop-and-report, if the valuation pull\n   failed or contradicted the item-count confirmation, from Step 3 and\n   Step 4. Or the full variance-by-item table from Step 5.\n4. The unmatched items from Step 6, if the reconciliation ran\n5. The total variance in the QBO company's actual home currency, if the\n   reconciliation ran. It covers every item that appears in both sources\n   and has a derivable cost basis. It leaves out items marked\n   \"unavailable\" per Step 5, and the two single-source unmatched\n   categories per Step 6.\n\n## What this skill never does\n\n- It never calls a `create_*`, `update_*`, or `delete_*` tool on the\n  QuickBooks Online MCP.\n- It never adjusts an inventory item's quantity or value in QBO.\n- It never reports \"no inventory tracked\" from an empty valuation pull\n  alone. It confirms with a separate item-count query first. It stops and\n  reports the failure, rather than skipping, when the valuation pull\n  itself errors, times out, or comes back malformed.\n- It never reports a variance for a business genuinely confirmed to track\n  no inventory in QBO. It states plainly that there is nothing to\n  reconcile, and it stops.\n- It never treats a genuinely empty count-against-QBO comparison as \"no\n  inventory tracked\". That comparison is a real zero variance. It\n  reconciles and reports the former. It skips only the latter.\n- It never guesses a per-unit cost when QBO's own on-hand quantity is\n  zero. It reports the dollar variance as unavailable instead.\n- It never does photo-based counting or OCR counting. The physical count\n  must already exist as a quantity list, pasted or uploaded.\n\n## Eval Contract\n\n### Spec\n\nA correct run compares a bookkeeper-supplied physical count against QuickBooks Online's own inventory records for the same period and reports the variance item by item, quantity and dollars, with each item classified as Matched, Short, Over, or Flagged. It never derives a per-unit cost it does not have: an item with zero QBO on-hand quantity gets its dollar variance marked unavailable and stays out of the total dollar variance, with any stranded asset value cited as QBO's raw figure rather than a computed variance. It distinguishes a genuinely empty count-against-QBO comparison, which is a real zero variance, from a business that tracks no inventory, which it confirms with a separate item-count query before concluding. It changes nothing in QuickBooks Online.\n\n### Rubric\n\nScore each dimension 0 or 1, total out of 6. Run the hard-fail gate first.\n\nThe hard-fail gate is evaluated over the run transcript and the output together, because a write call is a transcript fact. The scored table below is judged from the output alone. A condition appears in the gate or in the table, never in both.\n\n**Hard-fail gate (check before scoring):** Any call to a `create_*`, `update_*`, or `delete_*` tool on the QuickBooks Online MCP fails the run regardless of total, as does any claim to have adjusted an inventory item's quantity or value. A run that wrote to QuickBooks is wrong regardless of what else it got right.\n\n| # | Dimension | Pass | Fail | Weight |\n|---|-----------|------|------|--------|\n| 1 | No guessed cost basis | An item with zero QBO on-hand quantity has its dollar variance marked unavailable | A per-unit cost taken from another item, an average, or a round number | 1 |\n| 2 | Unavailable items excluded from the total | Any item marked unavailable is left out of the total dollar variance | Such an item folded into the total | 1 |\n| 3 | Stranded value flagged, not computed | Value-without-quantity cases classified Flagged, citing QBO's raw stranded asset value | Classified Matched because quantities agree, or reported as a computed variance | 1 |\n| 4 | Empty comparison is not \"no inventory\" | A real zero variance is reported as reconciled; only a confirmed no-inventory business stops the run | A zero variance reported as \"no inventory tracked\" | 1 |\n| 5 | No-inventory confirmed separately | \"No inventory tracked\" concluded only after a separate item-count confirmation | Concluded from an empty valuation pull alone | 1 |\n| 6 | Failed pull stops the run | A valuation pull that errors, times out, or returns malformed data is reported as a failure | A failed pull treated as an empty result and skipped | 1 |\n\n**Score to action:** 6\u002F6 ship. 5 acceptable, note the gap. 3 to 4 borderline, flag for human review. 0 to 2 bad, root-cause. Any hard-fail gate trip is a fail regardless of total.\n\n### Self-Test\n\n**Scenario A.** Period end 2026-06-30, single currency, count taken at end of day.\n\nPasted physical count:\n- Widget A: 100\n- Widget B: 40\n- Widget C: 0\n\nQuickBooks Online inventory records for the same date:\n- Widget A: 105 on hand, $12.00 per unit\n- Widget B: 40 on hand, $8.00 per unit\n- Widget C: 0 on hand, $250.00 inventory asset value\n\n- The output MUST classify Widget A as Short with a quantity variance of -5 and a dollar variance of -$60.00.\n- The output MUST classify Widget B as Matched with a zero quantity variance and a zero dollar variance.\n- The output MUST mark Widget C's dollar variance unavailable for lack of a cost basis and classify it as Flagged.\n- The output MUST cite QBO's raw $250.00 value for Widget C rather than reporting it as a computed variance.\n- The output MUST report a total dollar variance of -$60.00.\n- The output MUST NOT include Widget C's $250.00 in the total dollar variance.\n- The output MUST NOT derive a per-unit cost for Widget C from Widget A, from Widget B, or from an average.\n- The output MUST NOT call any `create_*`, `update_*`, or `delete_*` tool, or state that it adjusted a quantity.\n\n**Scenario B.** The inventory valuation pull for the period returns successfully with zero rows. The separate item-count query returns 0 inventory-type items.\n\n- The output MUST state that concluding \"no inventory tracked\" requires a separate item-count confirmation, and MUST report that confirmation's result before drawing that conclusion.\n- The output MUST distinguish this case from a count-against-QBO comparison that produced a genuine zero variance.\n- The output MUST NOT report \"no inventory tracked\" on the strength of the empty valuation pull alone.\n- The output MUST NOT report a reconciled zero variance as though a count had been compared against QBO records.\n\n### Version\n\n1.0.0\n\n---\n\n**More from Skills and Agents Co:** see this skill in the [Skills & Agents catalog](https:\u002F\u002Fskillsandagents.co\u002Fskills\u002Fqbo-inventory-reconciliation\u002F).\n",{"name":5,"license":15,"description":24},"Checks a physical inventory count against what QuickBooks Online thinks is on the shelf. It takes a per-item quantity list that the bookkeeper pastes or uploads, pulls QBO's inventory asset records for the same period through the QuickBooks Online MCP (intuit\u002Fquickbooks-online-mcp-server), and reports the variance item by item instead of as one lump number. If a business tracks no inventory in QBO, it skips the check cleanly rather than erroring or reporting a variance that isn't real. It writes nothing to QBO. It's read-only by design. Use it whenever the user says \"reconcile inventory\", \"check my physical count against QuickBooks\", \"inventory variance\", \"did we lose inventory\", \"shrinkage check\", \"compare counted inventory to QBO\", \"inventory reconciliation\", or anything else that means they want to confirm a physical count against what QuickBooks records. Always use this skill for QBO inventory reconciliation work. Don't freehand a variance comparison without it.\n","main","04d144d7a789a4f02020e70d10301dfb539c942f","plugins\u002Fquickbooks-close\u002Fskills\u002Fqbo-inventory-reconciliation",0,"https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library","https:\u002F\u002Fskillsandagents.co","https:\u002F\u002Fgithub.com\u002Fskills-agents-co","2026-11-12 05:58:29",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add skills-agents-co\u002Fskills-and-agents-library --skill qbo-inventory-reconciliation",{"tool":39,"command":40},"bunx","bunx skills add skills-agents-co\u002Fskills-and-agents-library --skill qbo-inventory-reconciliation",{"tool":42,"command":43},"pnpm","pnpm dlx skills add skills-agents-co\u002Fskills-and-agents-library --skill qbo-inventory-reconciliation","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fplugins\u002Fquickbooks-close\u002Fskills\u002Fqbo-inventory-reconciliation\u002FSKILL.md",true,[],[48,55,60,65,70,75],{"skillNo":49,"name":50,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":51,"categories":52,"keywords":53,"publishedAt":54},"01M76C3HSBFJ3RTPNEFFKB7KX0","x","https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fscripts\u002Ftest-fixtures\u002Fcheck-marketplace-parity\u002Fplugin-only-fail\u002Fplugins\u002Fq\u002Fskills\u002Fgamma\u002FSKILL.md",[],[],"2026-11-12 13:55:43",{"skillNo":56,"name":50,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":54},"01M76C3FZDA324S9SC2X5AZE0T","https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fscripts\u002Ftest-fixtures\u002Fcheck-marketplace-parity\u002Fplugin-only-fail\u002Falpha\u002FSKILL.md",[],[],{"skillNo":61,"name":50,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":62,"categories":63,"keywords":64,"publishedAt":54},"01M76C3F5EFJ69K4ZD6D0M2S76","https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fscripts\u002Ftest-fixtures\u002Fcheck-marketplace-parity\u002Findex-only-fail\u002Fbeta\u002FSKILL.md",[],[],{"skillNo":66,"name":50,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":54},"01M76C3EAP06YGXFT817A0KHW4","https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fscripts\u002Ftest-fixtures\u002Fcheck-marketplace-parity\u002Findex-only-fail\u002Falpha\u002FSKILL.md",[],[],{"skillNo":71,"name":50,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":54},"01M76C3CW1JMBTH9AYDP3MDWMS","https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fscripts\u002Ftest-fixtures\u002Fcheck-marketplace-parity\u002Ffinancial-pulse-shape-pass\u002Fplugins\u002Fextra\u002Fskills\u002Ffp-core\u002FSKILL.md",[],[],{"skillNo":76,"name":50,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":54},"01M76C3C2JQFW5GEF95TDKBTP8","https:\u002F\u002Fgithub.com\u002Fskills-agents-co\u002Fskills-and-agents-library\u002Fblob\u002Fmain\u002Fscripts\u002Ftest-fixtures\u002Fcheck-marketplace-parity\u002Ffinancial-pulse-shape-pass\u002Ffinancial-pulse\u002Fskills\u002Ffp-core\u002FSKILL.md",[],[]]