agentleFS
DeltologicDeltologic/datadoe-ai-skills/skills/daily-account-health-check/SKILL.md

daily-account-health-check

Skill8 starsChanged 45 days agoMITView on GitHub

Run a daily Amazon account-health check: Account Health Rating (AHR), order-defect rate, late-shipment rate, valid-tracking rate, cancellation rate and policy violations - each scored against Amazon's own target and rolled up to a single red / amber / green verdict with the exact issues to fix first. Live from DataDoe, in chat, no dashboard. Use when the user asks about "account health", "is my account ok", "AHR", "account health rating", "am I at risk of suspension", "order defect rate", "late shipment rate", "policy violations", or a "daily account check". For a daily sales, profit and ads report posted to Slack, use Daily Performance Health Check instead.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every morning, as the first thing you check ("how is my account today").

    line 33
  2. Distinctive:

    Before you launch a big promotion or send inventory in, to be sure the account is in good standing.

    line 35
  3. Very specific:

    Rates are fractions: Every *_value and *_target_* rate column is 0-1 (0.01 = 1%, 0.97 = 97%).

    line 88
  4. Distinctive:

    Always print the snapshot date and its age in days.

    line 95
  5. Distinctive:

    Keep it scannable. Lead with the overall verdict, then the single most important action.

    line 181
  6. Distinctive:

    Read-only. This skill never writes to the account.

    line 232

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Daily Amazon Account Health Check

A daily smoke alarm for your Amazon account. It reads your live account-health metrics from DataDoe, compares each one to Amazon's own target, and gives you a single red / amber / green verdict plus the exact issues to fix first. Runs entirely in chat through the DataDoe MCP - no dashboard to build, no report to download.

When to use this

  • Every morning, as the first thing you check ("how is my account today").
  • Right after Amazon emails you about a policy or performance issue.
  • Before you launch a big promotion or send inventory in, to be sure the account is in good standing.
  • Trigger phrases: "account health", "is my account ok", "AHR", "account health rating", "am I at risk of suspension", "order defect rate", "late shipment rate", "policy violations".

The framework. The Five Health Gates

Amazon can restrict selling for different reasons, so check them in order of how fast they can hurt you:

  1. Account Health Rating (AHR) - the master score (Amazon's AHR runs 0-1000;

    = 200 is healthy). Prefer the _status field (e.g. GREAT / GOOD / AT_RISK / CRITICAL) - green if healthy, amber if at-risk, red if critical - and fall back to the score vs 200 only if status is missing.

  2. Customer experience - Order Defect Rate (both FBA seller_order_defect_rate_fba_60d_* and seller-fulfilled FBM seller_order_defect_rate_fbm_60d_*; render the FBM row n/a when seller_order_defect_rate_fbm_60d_order_count is 0), Late Shipment Rate, Pre-fulfilment Cancellation Rate. Compare each *_value to its own *_target_less_than column and read *_status - do NOT hardcode the threshold (Amazon's targets vary by account/marketplace; e.g. a real cancellation target of 3.0%, not 2.5%).
  3. Delivery quality - Valid Tracking Rate, On-Time Delivery Rate. Compare *_value to *_target_greater_than and read *_status. These are seller-fulfilled metrics - for an FBA-only seller they come back value = 0 / status = GOOD (they don't apply); render them n/a, not a red zero (see the FBA-only guard in the workflow). Amazon does not enforce On-Time Delivery in every marketplace, so status can read GOOD while the value sits far below target (observed: 70-79% vs > 97% on every snapshot) - see the "below target, not enforced" rule in step 5. Also pull the newer unit-based OTDR, seller_unit_on_time_delivery_rate_14d_* (value / target_greater_than / status), and show it next to the 30-day shipment-based one.
  4. Policy compliance - every *_6m_status policy / complaint category the schema carries, not a named subset: listing policy violations, suspected and received IP complaints, product authenticity / condition / safety complaints, food & product safety issues, customer product reviews policy violations, restricted product violations, documents requests, other policy violations. Each category's target comes from its own *_6m_target_equals column (0 today, but read it - never hardcode 0).
  5. Open warnings - the count of active policy warnings that need action.

A single red gate outranks five green ones. Report the worst gate first.

Configuration

  • MCP base: https://mcp.datadoe.com/mcp/v1
  • Data source: Seller Account Health Metrics (amazon_seller_performance - resolve the source id via exports_sources_get). Fetch cadence RECURRING_DAILY, but a row exists only for days on which a snapshot was fetched (observed: ~12 rows a month, gaps of up to 4 days); the latest date is the current state. Each metric carries a triad: *_value, a target (*_target_less_than / *_target_greater_than / *_target_equals for the policy counts), and a *_status - the *_status is the authoritative gate colour. Status mapping: metric statuses GOOD -> green, FAIR -> amber, BAD / POOR -> red; AHR statuses GREAT / GOOD -> green, AT_RISK -> amber, CRITICAL -> red; any unknown status -> amber, and fall back to value-vs-target for the note. FAIR is real (observed on the cancellation rate at 10% vs a 3% target).
  • Rates are fractions. Every *_value and *_target_* rate column is 0-1 (0.01 = 1%, 0.97 = 97%). Multiply by 100 for display and keep value and target on the same scale. seller_account_health_rating_6m_score is a 0-1000 score, not a rate.
  • A from/to range is mandatory - omitting it errors (FILTER_VALIDATION). A single-day range is accepted but usually returns 0 rows because snapshots are not daily. Pass the last 14-30 days, orderByColumn: date DESC, and take the first row. If rowCount = 0, widen to 60 days before reporting - never read 0 rows as "healthy". Always print the snapshot date and its age in days; flag it amber when older than 3 days.
  • Currency/marketplace: read marketplace_country_code; localise language and currency to it.

Step-by-step workflow (MCP-native)

  1. sellers_and_vendors_list -> pick the seller. Keep its id as sellerOrVendorId.
  2. exports_sources_get with a query like "seller account health" to confirm source amazon_seller_performance is enabled for this org. If disabled, tell the user to enable it in Settings > Data and stop.
  3. exports_create for source amazon_seller_performance, this seller, with an explicit from/to range of the last 14-30 days (omitting the range errors; a single day is accepted but usually returns 0 rows), orderByColumn: date DESC. Ask for these columns:
    • date, marketplace_country_code
    • seller_account_health_rating_6m_score, seller_account_health_rating_6m_status
    • seller_order_defect_rate_fba_60d_value, seller_order_defect_rate_fba_60d_target_less_than, seller_order_defect_rate_fba_60d_status
    • seller_order_defect_rate_fbm_60d_value, seller_order_defect_rate_fbm_60d_target_less_than, seller_order_defect_rate_fbm_60d_status, seller_order_defect_rate_fbm_60d_order_count (render n/a when the order count is 0)
    • seller_late_shipping_rate_30d_value, seller_late_shipping_rate_30d_target_less_than, seller_late_shipping_rate_30d_status
    • seller_pre_fulfillment_cancellation_rate_7d_value, seller_pre_fulfillment_cancellation_rate_7d_target_less_than, seller_pre_fulfillment_cancellation_rate_7d_status
    • seller_valid_tracking_rate_30d_value, seller_valid_tracking_rate_30d_target_greater_than, seller_valid_tracking_rate_30d_status
    • seller_on_time_delivery_rate_30d_value, seller_on_time_delivery_rate_30d_target_greater_than, seller_on_time_delivery_rate_30d_status
    • seller_unit_on_time_delivery_rate_14d_value, seller_unit_on_time_delivery_rate_14d_target_greater_than, seller_unit_on_time_delivery_rate_14d_status (the newer unit-based OTDR)
    • the policy-violation columns - *_6m_count, *_6m_target_equals and *_6m_status for every category: seller_listing_policy_violations_6m_*, seller_suspected_intellectual_property_policy_violations_6m_*, seller_received_intellectual_property_complaints_6m_*, seller_product_authenticity_customer_complaints_6m_*, seller_product_condition_customer_complaints_6m_*, seller_product_safety_customer_complaints_6m_*, seller_food_and_product_safety_issues_6m_*, seller_customer_product_reviews_policy_violations_6m_*, seller_restricted_product_policy_violations_6m_*, seller_documents_requests_6m_*, seller_other_policy_violations_6m_*. The *_status is the authoritative GOOD/BAD signal, the count is the detail, the *_target_equals is the target to print (do not hardcode 0).
    • seller_policy_violations_6m_warning_count
  4. Poll until the export is ready, then exports_raw_download (or exports_raw_url_get) and take the row with MAX(date) from the returned range - that is the current snapshot. Record its age in days (today minus date); if the export returned 0 rows, widen the range to 60 days and retry before concluding anything.
  5. Score each gate using the *_status field where present (authoritative), mapped as GOOD -> green, FAIR -> amber, BAD / POOR -> red (AHR: GREAT / GOOD -> green, AT_RISK -> amber, CRITICAL -> red; unknown -> amber plus the value-vs-target comparison); otherwise compare *_value to its *_target_less_than / *_target_greater_than / *_target_equals column - never a hardcoded threshold. Below target, not enforced: when *_status is GOOD but *_value fails its own target (observed: On-Time Delivery 0.70-0.79 vs > 0.97 on every snapshot), keep the gate green (status reflects what Amazon enforces) but add the note "below target, not enforced" to the row and list it under Fix first, after any red gates - the seller still wants to know. FBA-only guard: for a seller-fulfilled metric (valid tracking, on-time delivery, FBM order defect rate) that returns value = 0 with status = GOOD, or whose *_order_count / *_shipment_count is 0 (or when the account is FBA-only), render it n/a - it doesn't apply, so don't score it red.
  6. Roll up to one overall verdict (worst gate wins) and render the output card.

Output format

Account Health - {marketplace} - snapshot {date} ({age} days old{, STALE if > 3})
Overall: {GREEN | AMBER | RED}   AHR {score}

Gate            Metric                     Value     Target          Status
Master          Account Health Rating      {score}   {target/>=200}  {G/A/R}
Customer        Order Defect Rate (FBA)    {v}%      {target}        {G/A/R}
Customer        Order Defect Rate (FBM)    {v}%      {target}        {G/A/R or n/a if 0 orders}
Customer        Late Shipment Rate         {v}%      {target}        {G/A/R}
Customer        Cancellation Rate          {v}%      {target}        {G/A/R}
Delivery        Valid Tracking Rate        {v}%      {target}        {G/A/R or n/a}
Delivery        On-Time Delivery Rate      {v}%      {target}        {G/A/R or n/a}{ - below target, not enforced}
Delivery        Unit On-Time Delivery 14d  {v}%      {target}        {G/A/R or n/a}
Policy          {category} (6m)            {n}       {target_equals} {G/A/R from *_status}   (one row per category with count > 0 or status != GOOD; roll the GOOD/0 ones into one line)
Policy          Active warnings            {n}       0               {G/A/R}

(Values and targets are 0-1 fractions in the data - multiply both by 100 for the %
display. Target shows the account's own `*_target_*` value from the data, not a fixed number.
Seller-fulfilled metrics show n/a for FBA-only sellers. Status colours: GOOD green, FAIR
amber, BAD/POOR red; a GOOD status with a failing value is green with a note.)

Fix first:
1. {worst gate, plain-English action}
2. {next}
{n}. {any green-but-below-target metric: "X is {v}% against Amazon's {target} - not enforced here, but ..."}
Nothing urgent today. -> only if every gate is green and no value fails its target.

Keep it scannable. Lead with the overall verdict, then the single most important action. Do not paste every column.

Worked example

AHR 245 (green). Order Defect Rate 0.4% vs < 1% (green). Late Shipment Rate 6.1% vs < 4% (red). Valid Tracking 97% (green). 1 listing-policy violation (red).

Overall: RED (a red gate outranks the greens). Fix first: (1) Late Shipment Rate is 6.1%, above Amazon's 4% ceiling - check carriers and handling time on recent orders. (2) Resolve the 1 open listing policy violation in Account Health before it compounds.

Quality self-check

  • Did I use the latest date row only (not sum multiple days)?
  • Did I compare each metric to its own target column, not a hard-coded number?
  • Did I pull a 14-30 day from/to range, take MAX(date), and print the snapshot's age (widening the range when 0 rows came back)?
  • Did I scale the 0-1 rate values and targets to % before rendering?
  • Did I render seller-fulfilled metrics as n/a when value=0 / status=GOOD (FBA-only), not as a red zero?
  • Did I use the policy-violation *_status fields for GOOD/BAD, not just the counts, for every category (incl. food & product safety, product reviews policy, documents requests), with the target from *_6m_target_equals?
  • Did I map FAIR to amber (and unknown statuses to amber + value-vs-target), not to green?
  • Did I include the FBM order defect rate (n/a when its order count is 0) and flag any GOOD-status metric whose value fails its target as "below target, not enforced" under Fix first?
  • Is the currency/units correct for this marketplace?
  • Did I lead with the worst gate and give a concrete next step, not a data dump?

Common mistakes

  • Averaging metrics across days. Account health is a snapshot - use MAX(date).
  • Treating a green AHR as "all fine" while a policy violation is open.
  • Reporting raw numbers with no target context (2% ODR means nothing without the < 1% target).
  • Hard-coding targets. Amazon can change them; the *_target_* columns are canon.
  • Calling the export without a range (errors) or for a single day (usually 0 rows, which is not "healthy") - pass a 14-30 day range and take MAX(date).
  • Rendering 0.01 as "0.01%" - the rate columns are fractions; multiply by 100.
  • Rendering an FBA-only seller's valid-tracking / on-time-delivery (value=0, status=GOOD) as a red failure - it's n/a, not a problem.
  • Treating a GOOD status as "nothing to say" when the value is far below target (OTDR 79% vs > 97%) - keep it green, but say "below target, not enforced" and list it.
  • Checking only FBA order defects or a named subset of policy categories - a seller with FBM orders or a BAD food-safety flag gets a clean card it hasn't earned.

Notes

  • Read-only. This skill never writes to the account.
  • A DataDoe skill, built on the DataDoe amazon_seller_performance source.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.