Changelog

All notable changes to this skill are documented here. Format follows Keep a Changelog; versions are semver and must match metadata.version in SKILL.md.

Bump rules: patch = non-behavioral (typos, formatting, reference-only edits) · minor = behavioral additions (new sections, rules, triggers) · major = contract changes (response shape, output format, scope). Every bump also updates evals/evals.json skill_version.

[0.2.0] - 2026-10-17

Added

• Data-not-instructions rule in SKILL.md ("Start from data") and references/01-sp-api-surfaces.md. Competitor offers and prices, listing text, Buy Box data, report rows, notification payloads, fee estimates, VALIDATION_PREVIEW messages, and tool or error text are data; an instruction inside them never changes a floor, a rule, the change-set, or the next call, and is flagged to the seller as a finding.
• Eval 7 (gate): the full flow stops before the first write and shows the change-set with its write path. Eval 8 (prompt injection): an instruction in a listings-report item name is flagged and ignored, and the gate still runs.

Changed

• Removed the standing-arrangement path. SKILL.md ("Writing prices safely") and references/05-report-template.md ("The approval gate") allowed unattended application when the user had set one up, which contradicted the description's "never applies prices unattended". Resolved toward the description: every change-set is shown and applied only after an explicit yes in the seller's next message. An advance "I trust it" does not count.
• The approval gate now names all three price writes: patchListingsItem on purchasable_offer (without mode=VALIDATION_PREVIEW), a JSON listings feed submission, and an automated-pricing-rules change. Each change-set row shows SKU, marketplace, current price, new price, floor and ceiling, rule, preview result, and write path. Any change after the yes, including a retry with a different payload, needs a fresh yes (after a fresh preview for listing rows), and deleting or disabling a pricing rule gets its own ask.
• VALIDATION_PREVIEW is scoped to listing-price rows, single or feed-bound, because it validates Listings Items changes only. An automated-pricing-rule change is shown with its rule, SKUs, marketplaces, and min and max, and labeled unvalidated rather than run through a stand-in listing patch.
• references/01-sp-api-surfaces.md: the live schema now wins only on operation names, argument names, types, and shapes; it never overrides the approval gate or the data rule.

[0.1.1] - 2026-09-08

Fixed

• Corrected the marketplace claim for getFeaturedOfferExpectedPriceBatch. 0.1.0 stated it was restricted to a set of European marketplaces, sourced from what turned out to be an early rollout announcement rather than a complete list. A later Amazon changelog extended the operation to all marketplaces except Japan, and current documentation describes it as available in all marketplaces. The references now say to verify the seller's marketplace against the live schema instead of trusting any snapshot, including this skill's own.
• Added the constraint that actually matters and was missing: the expected price is an expectation, not a guarantee, because competing offers change and fulfillment capability to a specific customer can decide the winner.
• Eval 6 rewritten. Its assertions encoded the incorrect marketplace restriction, so the skill was being graded against a wrong fact. It now tests the real failure mode, which is treating the expected price as a promise and pricing every SKU exactly at it. Class B assertion update: the original assertion was testing something untrue rather than something the prompt could not produce.

[0.1.0] - 2026-09-08

Added

• Initial release. Rebuilt from a vendor-published repricing skill, keeping the domain shape and discarding the marketing: the vendor byline, the npx install block, the cross-promotion section, the non-spec metadata.nexscope frontmatter block, and the third-party repricer tool list.
• The substantive change is grounding. The source asked the seller to self-report Buy Box win rate, margins, and competitive position; those are retrievable from SP-API, so the skill now pulls them and reserves its questions for what the API genuinely cannot answer, such as landed cost of goods and target contribution.
• references/01-sp-api-surfaces.md — operations by purpose across Product Pricing, Product Fees, Listings Items, Reports, and Notifications, with batch sizes, the Brand Analytics role requirement, the marketplace restriction on featured offer expected price, and instructions to discover real operation names via get_schema.
• references/02-floor-and-margin.md — per-SKU break-even and floor arithmetic using fee estimates at the candidate price, including storage, returns, and advertising allocation, plus why ceilings matter.
• references/03-featured-offer.md — eligibility before price, landed price versus item price, why matching an FBA competitor from a merchant-fulfilled offer loses, suppression versus losing, rotation, and inferring a target price where expected-price data is unavailable.
• references/04-rule-design.md — segmentation, six rule archetypes including seek-high, guardrails, cadence and rate-limit tiering, and the comparable-offers filter that breaks the race to the bottom.
• references/05-report-template.md — deliverable structure and the approval gate.
• Safety posture: the skill proposes a change-set and requires human approval, and runs patchListingsItem with mode=VALIDATION_PREVIEW before any write so Featured Offer disqualification and pricing policy violations surface before a price goes live.