Agent Skills: HubSpot Deal Pipeline Automation

HubSpot Deal Pipeline Automation

Overview

Automate HubSpot deal pipelines without blowing up real revenue. This is not a walkthrough of workflow builder UI — it is the engineering behind six failure modes that silently destroy RevOps data while every dashboard stays green until a rep's quota call reveals the damage.

The six production failures this skill prevents:

1. Stage automation loops — a workflow moves a deal to "Demo Scheduled", which triggers a second workflow enrolled on "Demo Scheduled" that moves it back to "Qualified", which triggers the first workflow again. The deal bounces between stages until HubSpot's 100-action daily workflow execution limit for that deal is exhausted. RevOps sees "workflow failure" alerts but no stage history that makes sense.

2. Stale-deal auto-closing — a cleanup workflow auto-closes deals older than 90 days to clear pipeline bloat. The contact associated with deal #4872 responded to an email this morning. The deal is closed as Lost. The rep calls the contact, who says "I just told someone we were ready to sign." This failure costs quota attainment, not just data quality.

3. Forecast query inconsistency — amount, hs_projected_amount, and a custom arr_value property all exist on the same deal. The CRO dashboard queries amount. The ops team queries hs_projected_amount. RevOps built a custom rollup on arr_value. At end of quarter they produce three different forecast numbers for the same deal set with no canonical answer.

4. Custom property reporting gap — a RevOps engineer renames deal_source_detail to original_lead_source in the property settings. Every deal-source dashboard built on the old property name silently returns zero. No error is thrown. The board sees a 100% collapse in deal source tracking that is entirely an artifact, not a pipeline signal.

5. Quota dashboard staleness — HubSpot's reporting layer caches deal stage aggregates. A deal closes at 4:47pm. The quota attainment dashboard doesn't reflect it until 8:30pm. A rep who hit quota at EOD is told they're 3% short. This is a 4-hour cache lag in the default reporting stack, and it affects every quota conversation at month-end.

6. Multi-pipeline deal duplication — a new business opportunity becomes an expansion after the initial close. Someone creates a deal in the Expansion pipeline for the same company without deleting the original, or a workflow auto-creates expansion deals. The same revenue is counted in two pipeline forecasts with no cross-reference key linking them.

Prerequisites

All API calls use a HubSpot private app token or OAuth access token in the Authorization: Bearer $HUBSPOT_TOKEN header. The hubspot-auth skill covers token caching, rotation, and rate-limit backoff patterns. For deal pipeline automation, the required scopes are listed below.

• HubSpot Sales Hub Professional or Enterprise (workflows + custom properties
  • API access)
• Private app token with scopes: crm.objects.deals.read, crm.objects.deals.write, crm.schemas.deals.read, crm.schemas.deals.write, crm.associations.read, crm.associations.write, automation
• Node.js 18+ or Python 3.10+ in the execution environment
• jq installed for shell-based audit scripts
• HubSpot portal ID (visible at top-right of any HubSpot page, or via GET /oauth/v1/access-tokens/$TOKEN)

Instructions

Build in this order. Each section neutralizes one production failure mode.

1. Stage transition guard (neutralizes automation loops)

Workflow loops happen when a stage-change trigger has no memory of what caused the change. The guard injects a sentinel property hs_pipeline_loop_guard (type: string, format: workflowId:epochMs) that any workflow writes immediately before it changes a stage, and reads before it fires. If the guard was written by the same workflow within a debounce window, the workflow aborts.

Create the sentinel property first (one-time setup per portal):

curl -s -X POST "https://api.hubapi.com/crm/v3/properties/deals" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "hs_pipeline_loop_guard",
    "label": "Pipeline Loop Guard",
    "type": "string",
    "fieldType": "text",
    "groupName": "dealinformation",
    "description": "Automation loop sentinel. Format: workflowId:epochMs. Written before any stage transition."
  }' | jq '{name, label, type}'

TypeScript guard — wrap every stage-change call in this:

const DEBOUNCE_MS = 60_000; // 60s window per workflow

async function safeStageTransition(
  dealId: string,
  workflowId: string,
  targetStageId: string,
  token: string,
): Promise<{ transitioned: boolean; reason: string }> {
  // Read current guard value
  const deal = await hubspotGet(`/crm/v3/objects/deals/${dealId}`, token, {
    properties: "dealstage,hs_pipeline_loop_guard",
  });

  const guard: string | null = deal.properties.hs_pipeline_loop_guard ?? null;
  if (guard) {
    const [guardedWorkflowId, tsStr] = guard.split(":");
    const elapsed = Date.now() - parseInt(tsStr, 10);
    if (guardedWorkflowId === workflowId && elapsed < DEBOUNCE_MS) {
      return {
        transitioned: false,
        reason: `Loop guard: workflow ${workflowId} already fired ${elapsed}ms ago`,
      };
    }
  }

  // Write guard, then transition
  await hubspotPatch(`/crm/v3/objects/deals/${dealId}`, token, {
    properties: {
      hs_pipeline_loop_guard: `${workflowId}:${Date.now()}`,
      dealstage: targetStageId,
    },
  });

  return { transitioned: true, reason: "OK" };
}

hubspotGet and hubspotPatch are thin fetch wrappers — implementations in implementation-guide.md.

Detecting existing loops via stage history audit:

# Pull stage change history for a specific deal — look for oscillation
DEAL_ID="12345678"
curl -s "https://api.hubapi.com/crm/v3/objects/deals/${DEAL_ID}/changelog" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" | \
  jq '.results[] | select(.propertyName == "dealstage") | {timestamp, from: .previousValue, to: .currentValue}'

If the output shows the same two stage IDs alternating more than twice in 24 hours, that deal has a live loop. Identify the responsible workflows by checking hs_lastmodifieddate on the deal alongside the workflow enrollment history in HubSpot UI (Contacts → Workflows → History tab for the deal).

2. Stale-deal safe-close audit (neutralizes stale-deal closing)

Never auto-close a deal without first checking whether the associated contact has recent email activity. The safe-close criterion is: deal older than 90 days AND no inbound email in the last 30 days AND no open tasks.

The full stale_deal_audit.py script (with get_stale_open_deals, has_recent_inbound_email, has_open_tasks, and paginated search) lives in implementation-guide.md § Stale-Deal Audit.

Key points for the search query:

# CRM search filter for stale open deals (epoch ms cutoff)
CUTOFF=$(python3 -c "import datetime; print(int((datetime.datetime.utcnow()-datetime.timedelta(days=90)).timestamp()*1000))")
# filterGroups: createdate LT $CUTOFF AND hs_is_closed EQ false

The script outputs a CSV with safe_to_close boolean. Only rows where safe_to_close=True are candidates for batch close. A separate batch_close_stale.py script (also in the implementation guide) handles the actual close — always requires human review of the CSV before executing.

3. Forecast amount reconciliation (neutralizes query inconsistency)

HubSpot exposes four distinct amount-like fields on a deal. Using the wrong one in a report produces a different number without any error. The canonical selection logic:

| Field | When it is the right number | |---|---| | amount | The rep-entered deal value. Use for quota credit and pipeline value in all CRM views. | | hs_projected_amount | Amount × current-stage probability. Use only for probability-weighted pipeline forecasts. | | hs_deal_stage_probability | The stage probability (0–1). Multiply amount by this yourself if you need weighted values. | | Custom arr_value, mrr_value, etc. | Only if your org has explicitly separated ARR/MRR from deal amount. Validate these exist on the deal before reading them. |

Always request all four amount fields — pick one canonical field per report:

# Fetch all four amount fields for a deal and compare
curl -s "https://api.hubapi.com/crm/v3/objects/deals/$DEAL_ID" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" \
  --data-urlencode "properties=amount,hs_projected_amount,hs_deal_stage_probability,arr_value" | \
  jq '.properties | {amount, projected: .hs_projected_amount, probability: .hs_deal_stage_probability, arr: .arr_value}'

Discrepancy rule: if |hs_projected_amount - (amount × hs_deal_stage_probability)| / amount > 0.05, hs_projected_amount is stale. Force recalculation by patching dealstage to the current stage value — HubSpot recomputes on every stage write.

The full Python forecast_reconcile.py script (with paginated open-deal search and --fix flag for bulk recalculation) is in implementation-guide.md § Forecast Reconciliation.

4. Custom property drift detection (neutralizes reporting gaps)

When a deal property used in a dashboard is renamed or deleted, dashboards silently return zero without throwing an error. Run this check before every sprint that touches property schema, and wire it into a nightly CI job.

#!/usr/bin/env bash
# check-deal-properties.sh
# Usage: HUBSPOT_TOKEN=... ./check-deal-properties.sh properties-in-use.txt
# properties-in-use.txt: one property name per line (from dashboard/report definitions)

set -euo pipefail

PROPERTIES_FILE="${1:-properties-in-use.txt}"
TOKEN="${HUBSPOT_TOKEN:?HUBSPOT_TOKEN required}"

# Fetch all deal properties from the portal
ALL_PROPS=$(
  curl -s "https://api.hubapi.com/crm/v3/properties/deals" \
    -H "Authorization: Bearer $TOKEN" | \
  jq -r '.results[].name'
)

echo "Checking properties against portal schema..."
MISSING=0

while IFS= read -r prop; do
  [[ -z "$prop" ]] && continue
  if ! echo "$ALL_PROPS" | grep -qx "$prop"; then
    echo "MISSING: $prop"
    MISSING=$((MISSING + 1))
  fi
done < "$PROPERTIES_FILE"

if [[ $MISSING -gt 0 ]]; then
  echo ""
  echo "ERROR: $MISSING properties used in reports do not exist in this portal."
  echo "Review recent property renames/deletions in Settings → Properties."
  exit 1
else
  echo "All $( wc -l < "$PROPERTIES_FILE" | tr -d ' ' ) properties found in portal schema."
fi

Generate properties-in-use.txt by extracting all property names from your report/dashboard definitions. The list must be maintained alongside any schema change — wire check-deal-properties.sh into your deployment pipeline so that property renames fail the deploy before they break dashboards in production.

5. Quota dashboard cache-busting (neutralizes staleness)

HubSpot's native reporting layer caches deal stage aggregates. The cache TTL in the standard reporting stack is up to 4 hours. For real-time quota views at month-end, bypass the reporting layer and query the CRM API directly with a current-state search.

interface QuotaSnapshot {
  closedWon: number;    // sum of amount for ClosedWon deals in current period
  closedWonCount: number;
  openPipeline: number; // sum of amount for open deals
  asOf: string;         // ISO timestamp of this query
}

async function liveQuotaSnapshot(
  pipelineId: string,
  closedWonStageId: string,
  ownerId: string,
  token: string,
  periodStartMs: number,
): Promise<QuotaSnapshot> {
  // Closed Won this period
  const closedWonSearch = {
    filterGroups: [
      {
        filters: [
          { propertyName: "pipeline", operator: "EQ", value: pipelineId },
          { propertyName: "dealstage", operator: "EQ", value: closedWonStageId },
          { propertyName: "hubspot_owner_id", operator: "EQ", value: ownerId },
          { propertyName: "closedate", operator: "GTE", value: String(periodStartMs) },
        ],
      },
    ],
    properties: ["amount"],
    limit: 100,
  };

  const cwResp = await hubspotSearch("/crm/v3/objects/deals/search", closedWonSearch, token);
  const closedWon = cwResp.results.reduce(
    (sum: number, d: any) => sum + (parseFloat(d.properties.amount) || 0),
    0,
  );

  // Open pipeline (not closed) — for forecast view
  const openSearch = {
    filterGroups: [
      {
        filters: [
          { propertyName: "pipeline", operator: "EQ", value: pipelineId },
          { propertyName: "hs_is_closed", operator: "EQ", value: "false" },
          { propertyName: "hubspot_owner_id", operator: "EQ", value: ownerId },
        ],
      },
    ],
    properties: ["amount"],
    limit: 100,
  };

  const openResp = await hubspotSearch("/crm/v3/objects/deals/search", openSearch, token);
  const openPipeline = openResp.results.reduce(
    (sum: number, d: any) => sum + (parseFloat(d.properties.amount) || 0),
    0,
  );

  return {
    closedWon,
    closedWonCount: cwResp.total,
    openPipeline,
    asOf: new Date().toISOString(),
  };
}

hubspotSearch is the same helper defined in section 1. Serve this function from your internal ops dashboard and poll it on page load instead of embedding a HubSpot report iframe. The CRM API reflects stage changes within seconds of them happening. The full polling service (Express + 30-second in-memory cache

• React widget) is in implementation-guide.md § Quota Dashboard Cache-Busting Pattern.

6. Multi-pipeline duplicate detection (neutralizes deal duplication)

The canonical deduplication key is company_id + closedate_month + deal_type_custom_field. Without this cross-reference, the same opportunity appears in two pipelines and inflates ARR forecast.

The full findDuplicateDeals() and searchAllOpenDeals() TypeScript implementations live in implementation-guide.md § Multi-Pipeline Duplicate Detection. The algorithm:

1. Search all open deals, paginated (hs_is_closed EQ false)
2. For each deal, fetch associated companies via GET /crm/v3/objects/deals/$DEAL_ID/associations/companies
3. Group deals by companyId — companies with deals in more than one pipeline are duplicates

Once duplicates are identified, link them with a cross-reference association before closing one:

# Link a new-business deal to its expansion deal (association type 5 = deal-to-deal)
curl -s -X PUT \
  "https://api.hubapi.com/crm/v4/objects/deals/${NB_DEAL_ID}/associations/deals/${EXP_DEAL_ID}/5" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" | jq '{fromObjectId, toObjectId, associationTypes}'

Error Handling

| HTTP Status | Error | Root Cause | Action | |---|---|---|---| | 400 BAD_REQUEST | INVALID_FILTER | Search filter uses a property name that doesn't exist or uses wrong operator for type | Validate property names via GET /crm/v3/properties/deals before building filter; use EQ for enum, GTE/LT for date/number | | 400 BAD_REQUEST | PROPERTY_DOESNT_EXIST | PATCH payload includes a property that was deleted or renamed | Run check-deal-properties.sh against the payload property list; catch and surface the missing property name to RevOps | | 403 FORBIDDEN | MISSING_SCOPES | Token lacks crm.objects.deals.write or automation scope | Verify scopes via GET /oauth/v1/access-tokens/$TOKEN; re-issue private app token with correct scopes | | 404 NOT_FOUND | OBJECT_NOT_FOUND | Deal ID in URL doesn't exist, or association target object doesn't exist | Confirm deal exists with GET /crm/v3/objects/deals/$DEAL_ID before writing; handle 404 in batch loops gracefully | | 409 CONFLICT | OBJECT_ALREADY_EXISTS | Batch create includes a deal that matches a unique property value | Deduplicate input set; use batch update instead of create for existing deals | | 429 TOO_MANY_REQUESTS | RATE_LIMIT | Exceeded 100 API calls/10s (Professional/Enterprise) or 250K/500K daily | Read Retry-After header and pause; add 100ms sleep between pages in search loops; use batch endpoints to collapse N updates into one call |

Examples

Search for all deals closing this month

MONTH_START=$(python3 -c "import datetime; now=datetime.datetime.utcnow(); print(int(datetime.datetime(now.year,now.month,1).timestamp()*1000))")
MONTH_END=$(python3 -c "import datetime; now=datetime.datetime.utcnow(); import calendar; end=calendar.monthrange(now.year,now.month)[1]; print(int(datetime.datetime(now.year,now.month,end,23,59,59).timestamp()*1000))")

curl -s -X POST "https://api.hubapi.com/crm/v3/objects/deals/search" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" \
  -H "Content-Type: application/json" \
  -d "{
    \"filterGroups\": [{
      \"filters\": [
        {\"propertyName\": \"closedate\", \"operator\": \"GTE\", \"value\": \"$MONTH_START\"},
        {\"propertyName\": \"closedate\", \"operator\": \"LTE\", \"value\": \"$MONTH_END\"},
        {\"propertyName\": \"hs_is_closed\", \"operator\": \"EQ\", \"value\": \"false\"}
      ]
    }],
    \"properties\": [\"dealname\",\"amount\",\"dealstage\",\"pipeline\",\"hubspot_owner_id\"],
    \"sorts\": [{\"propertyName\": \"amount\", \"direction\": \"DESCENDING\"}],
    \"limit\": 50
  }" | jq '.results[] | {id, name: .properties.dealname, amount: .properties.amount, stage: .properties.dealstage}'

Batch update deal stage for a list of deal IDs

# deals-to-update.json: {"id":"12345","properties":{"dealstage":"appointmentscheduled"}}
jq -n --argjson ids '["11111","22222","33333"]' \
  '{inputs: [$ids[] | {id: ., properties: {dealstage: "qualifiedtobuy"}}]}' \
  > batch_payload.json

curl -s -X POST "https://api.hubapi.com/crm/v3/objects/deals/batch/update" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" \
  -H "Content-Type: application/json" \
  -d @batch_payload.json | jq '{status, numResults: (.results | length), errors: (.errors // [])}'

List all pipelines and their stages

curl -s "https://api.hubapi.com/crm/v3/pipelines/deals" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" | \
  jq '.results[] | {pipelineId: .id, label: .label, stages: [.stages[] | {stageId: .id, label: .label, probability: .metadata.probability}]}'

Enumerate all custom deal properties

curl -s "https://api.hubapi.com/crm/v3/properties/deals" \
  -H "Authorization: Bearer $HUBSPOT_TOKEN" | \
  jq '[.results[] | select(.hubspotDefined == false) | {name, label, type, fieldType, groupName}]'

Output

Working with this skill produces:

• Stage transition guard — hs_pipeline_loop_guard sentinel property created in the portal, and a TypeScript safeStageTransition() wrapper that prevents any workflow from re-triggering itself within a 60-second window
• Stale-deal audit CSV — stale_deals.csv with a safe_to_close boolean column; only rows with safe_to_close=True are candidates for auto-close workflows
• Forecast reconciliation log — structured WARN output for every deal where hs_projected_amount diverges from amount × stage_probability by more than 5%, identifying deals where stage probability was changed without recalculating projections
• Property drift report — exit-1 CI gate that lists every report property not present in the portal's current schema
• Live quota snapshot endpoint — direct CRM API query returning closedWon, closedWonCount, openPipeline, and asOf with sub-second freshness versus the 4-hour reporting cache
• Duplicate deal groups — list of companies with open deals in more than one pipeline, with deal IDs to inspect and cross-reference associations to create before merging

Resources

• HubSpot CRM API — Deals
• CRM Search API
• Pipelines API
• Associations API v4
• Deal Properties Reference
• Workflow Enrollment History
• API_REFERENCE.md — pipeline/stage endpoint shapes, deal property catalog, search filter syntax, association types
• implementation-guide.md — stale-deal audit script, forecast reconciliation query, loop guard deployment, cache-busting quota dashboard pattern