E-commerce brands · EC-04
Returns to Corrective Action Brief
Reconcile return events to original units before deciding what to investigate or change.
$49 proposed one-time price · one named user
Paid release in preparation. No purchase is available yet. Prices exclude applicable taxes.
What this workflow is designed to do.
Review one sales cohort with up to ten SKUs and 100 supplied redacted return records, then draft traceable return metrics, investigations and proposed corrective actions.
Use the input and output contract above to assess whether it matches the decision you need to make. The examples below show the intended structure, with source information and human review kept visible.
SEE THE SHAPE OF THE WORK
From supplied facts to a useful draft.
Example input
# Fictional input — P300 May sales cohort Synthetic fixture only: no real customers, orders or outcomes. These batch-local keys identify fictional units, not people. [S1] Cohort MAY-P300 includes original orders placed May 1–31, 2026 inclusive, timezone Africa/Casablanca. Only SKU P300 is in scope. Follow-up cutoff: June 30, 2026 at 23:59:59 Africa/Casablanca. Event dates below use that timezone. Source owner Alex, product reviewer Lee, operations reviewer Mina. Alex confirms that the eight supplied rows cover all physical-return and linked status events through the cutoff; the export also contains the explicitly dated later event R07. No maturity-window definition is supplied. [D1] Approved fulfilled-unit source for this cohort and cutoff: 10 original units. Line L1: P300, ordered May 2, fulfilled May 3, five original units A1–A5. Line L2: P300, ordered May 30, fulfilled May 31, five original units B1–B5. Aggregate 10 = 5 + 5. Old line L0: P300, ordered April 30, fulfilled May 1, unit X1; outside the original-order-date cohort and outside the denominator. Replacement shipments and unfulfilled cancellations are excluded from D1. [T1] Alex-approved status definitions: RECEIVED confirms physical receipt on the event date; REFUNDED confirms a financial status only; EXCHANGE_REQUESTED is only a request. No receipt reversal is present. Stable source event IDs identify duplicates; duplicate R04 is confirmed below. Repeated statuses for the same original unit are history, not new units. [T2] Approved reason taxonomy: SIZE = stated fit concern; DAMAGE = stated arrival damage; CHANGED_MIND = stated preference change. Multiple simultaneous codes per unit are allowed. R01 and R02 carry the same confirmed current reason for A1; R03 is the confirmed current reason set for B1. A code does not establish the cause. [P1] Lee-approved facts for P300: fabric pouch with external dimensions 30 × 20 cm; hand wash only. No internal dimensions, sizing-error finding, packaging-damage investigation or corrective change is approved. Financial bases, action dates and an observation-window target are absent. Return/event rows, one unit per row: R01 — event E01; L1/A1; P300; May 20, 10:00; RECEIVED; reason SIZE; physical receipt confirmed by T1. R02 — event E02; L1/A1; P300; May 22, 10:00; REFUNDED; reason SIZE; status history for A1, no new physical return. R03 — event E03; L2/B1; P300; June 10, 10:00; RECEIVED; reasons SIZE and DAMAGE; physical receipt confirmed by T1. R04 — event E03; L2/B1; P300; June 10, 10:00; RECEIVED; reasons SIZE and DAMAGE; confirmed exact duplicate of R03. R05 — event E05; L1/A2; P300; May 28, 10:00; EXCHANGE_REQUESTED; reason SIZE; no physical-receipt evidence. R06 — event E06; L0/X1; P300; May 15, 10:00; RECEIVED; reason CHANGED_MIND; original order April 30 per D1. R07 — event E07; L2/B2; P300; July 2, 10:00; RECEIVED; reason SIZE; after the declared follow-up cutoff. R08 — event E08; L2/B3; P300; June 18, 10:00; REFUNDED; reason CHANGED_MIND; financial status only, no physical-receipt evidence. Decision requested: reconcile this cohort and draft investigations and proposed corrective actions for editorial/operations review. Do not contact customers or change products, processes or accounts.
Illustrative output
# MAY-P300 — returns and corrective-action brief Hand-authored synthetic illustration; not recorded model output. DRAFT — HUMAN REVIEW REQUIRED. Readiness: source-linked reconciliation available for review; no action has been approved. Reviewers: Lee and Mina; source owner Alex. Scope and sources Cohort: original P300 orders placed May 1–31, 2026, Africa/Casablanca; follow-up through June 30 at 23:59:59 in that timezone. S1 defines scope; D1 supplies original fulfilled-unit counts and membership; T1 defines receipt/status meaning; T2 defines reasons; P1 supplies approved product facts. Approval is supplied status, not independent verification. Record reconciliation 8 supplied rows = 1 duplicate + 1 out-of-cohort + 1 after-cutoff + 0 unmatched + 0 held + 5 linked evidence rows. Duplicate: R04 repeats R03, same event E03. Out-of-cohort: R06, original sale April 30 on L0; its May receipt does not make it a May sale. After-cutoff: R07, July 2 receipt for B2; not included in the June 30 result. Unmatched and held: none in this supplied fixture. Linked evidence: R01, R02, R03, R05 and R08. Unit reconciliation 2 known physical returned units: A1 from R01, and B1 from R03. R02 is A1's refund history and adds no returned unit. A2 has only exchange-request evidence R05; B3 has only refund evidence R08. Neither establishes physical receipt. R04, R06 and R07 add no unit to the confirmed in-scope count under the stated cutoff. Metrics and denominator distinction D1 supplies 10 original fulfilled units: L1 five plus L2 five. Old unit X1 is outside the sales cohort. The two confirmed returned units belong to D1. Cohort physical-return rate through cutoff: 2 distinct original units / 10 original fulfilled units = 20%. The numerator and denominator share S1/D1 cohort scope and cutoff; S1 supplies complete physical-return coverage. This is a descriptive cutoff result, not a final rate or a claim about other cohorts. Reason shares among the 2 known confirmed returned units: SIZE = A1 and B1, 2/2 = 100%; DAMAGE = B1, 1/2 = 50%; unknown reasons = 0/2. B1 has both codes, so reason shares sum to 150%. These are shares of known returned units, not rates among the 10 fulfilled units. A2's requested exchange and B3's refund-only record do not enter this reason denominator. Follow-up and interpretation The May sales cohort includes physical receipts in June, including B1 on June 10. B2's July receipt is outside the cutoff and signals why this snapshot is not final. Fulfillment dates span May 3–31, so units have unequal follow-up at June 30. Maturity is UNRESOLVED because no approved observation window is supplied. Two confirmed units support descriptive investigation only; they do not establish a product defect, warehouse fault or a causal pattern. Financial impact is not calculated; no approved financial bases are supplied. Investigation 1 — preserve unit/status separation Observation: R02 is additional history for A1, while R05 and R08 show no physical receipt [T1]. Proposed action: Alex reviews the event-to-original-unit mapping before this draft is used in reporting, retaining requests and financial statuses separately from received units. The present source mapping supports the stated counts; no missing unit is invented. Owner: Alex. Date: UNRESOLVED. Status: proposed review of the supplied reconciliation. Investigation 2 — examine fit information Observation: SIZE appears on A1 and B1, 2 of the 2 known returned units [R01, R03, T2]. This does not prove incorrect dimensions or justify redesign. Approved fact: P1 provides external dimensions 30 × 20 cm; no internal dimensions or sizing-error finding. Proposed action: Lee reviews whether approved sales copy clearly labels external dimensions and identifies what additional fit evidence is needed. Any clarification draft must use approved facts and receive review before use. Owner: Lee. Date: UNRESOLVED. Status: proposed; no product change or new compatibility claim approved. Investigation 3 — examine the damage record Observation: DAMAGE appears on B1, 1 of the 2 known returned units [R03, T2]. The code alone does not identify the cause. Proposed action: Mina reviews authorized condition and packaging evidence for B1 and records a finding before proposing a packaging or handling change. No such evidence is supplied here. Owner: Mina. Date: UNRESOLVED. Status: proposed; no carrier blame, replacement or process change approved. Human decision queue Alex: review the original-unit/status reconciliation and supplied coverage declaration, and define the intended observation window. Lee: review the fit-information investigation and any future source-backed copy proposal. Mina: review damage evidence and any resulting action proposal. Reviewer/date/decision: blank for human completion. No customer contact, refund, replacement, product/process change or external execution occurred.
Fictional, hand-authored reference example. This is not a customer result or an evaluated AI run. No accuracy or time-saving results have been established.
Inside the planned package.
The workflow
A SKILL.md instruction file, a quickstart, and installation guidance.
The working materials
Input and output templates, example documents, and a review checklist.
The boundaries
Evaluation cases, version history, license, and data-handling notes.
Before you get started.
Environment & installation
Claude Code is a candidate for the first pilot, not a verified supported host. Host choice, account requirements, and installation must be validated with pilot users. AI tool costs are separate from the proposed product price.
Read the installation previewRelease evidence
Prototype v0.1.0 · Last edited 2026-09-07. Authoring: LibSkills prototype project. Independent industry reviewer: not yet assigned. Host test date and result: none. See the required quality gate.
Limitations & review points
- Proposed paid product; no paid package is available yet.
- One declared sales cohort, at most ten SKUs and 100 supplied redacted records; no account access or customer contact.
- Missing or inconsistent unit denominators and unresolved numerator coverage block rates; small samples do not establish causes.
- Does not estimate lost revenue, issue refunds or implement product/process changes.
- Private authored candidate; installation, behavioral runs and user pilots remain pending.
Proposed paid terms
12 months of updates and 30 days of setup support. Client deliverable use allowed; redistribution of paid source excluded. License proposal · Refund proposal.
THE NEXT PIECE OF YOUR WORKFLOW
Keep the work moving.
Collection preview
E-commerce brands ↗
Multi-channel Product Fact QA
Find factual drift across two or three product drafts while preserving variants, units, components and pack quantities.
A claim ledger with exact draft locations, variant scope and source-field references
Collection preview
E-commerce brands ↗
Customer Feedback to Product Page Brief
Turn supplied customer questions into product-page changes backed by approved facts.
A row audit and overlapping objection themes with unique-excerpt denominators
Collection preview
E-commerce brands ↗
Seasonal Campaign Consistency Check
Catch mismatched discounts, thresholds, dates, asset versions and stock promises before a seasonal campaign is approved.
A source-linked campaign contract and draft inconsistency review