Crm Activity Sync

/crm-activity-sync — CRM Activity Sync for Solution Engineers

Read structured activity logs from customer-engagements and push them into MSX CRM (Dynamics 365) as milestone activities. This skill is the second stage of a two-stage pipeline:

/daily-activity-log → activity-log.md → /crm-activity-sync → MSX CRM tasks

/daily-activity-log collects evidence (git repos, calendar, emails) and writes activity-log.md in each project folder. This skill reads those logs, maps each entry to the correct MSX milestone, and creates CRM tasks.

Architecture: Pipeline Model

┌─────────────────────────┐     ┌──────────────────────────┐     ┌──────────┐
│  /daily-activity-log     │     │  customer-engagements/    │     │  MSX CRM │
│  (upstream skill)        │     │  (single source of truth) │     │          │
│                          │     │                           │     │          │
│  📦 Work repos (git)    │────▶│  activity-log.md          │     │          │
│  📅 Calendar events     │────▶│  meetings/                │────▶│  Tasks   │
│  📧 PG/Eng emails       │────▶│  followups.md             │     │          │
│  💬 Teams (future)      │────▶│  email-threads/ (future)  │     │          │
└─────────────────────────┘     └──────────────────────────┘     └──────────┘

 Runs daily at 07:30          Source of truth                  Runs daily at 08:00

This skill reads ONLY from customer-engagements/. It never directly queries calendar, git repos, or email. That is /daily-activity-log's job.

Platform Compatibility

This skill runs on macOS, Linux, and Windows. Detect the OS first and pick the right command syntax. See _shared/PLATFORM.md (skills repo root) for the full translation table. Quick reference:

VPN check note: ~/Scripts/ensure-vpn.sh is POSIX-only. On Windows, the user must connect Azure VPN Client manually before running this skill, or you must provide a PowerShell equivalent (ensure-vpn.ps1). If neither exists on Windows, ask the user to confirm VPN is connected and proceed.

Default to POSIX commands; fall back to PowerShell when running on Windows native (not WSL, not Git Bash).

Core Principles

• Single source of truth. All activity evidence lives in customer-engagements/. This skill reads activity-log.md — nothing else. If an activity isn't logged there, it doesn't get synced to CRM.
• Never fabricate activities. Only create CRM tasks backed by entries in activity-log.md. If the log is empty for a date, create nothing.
• Idempotent per daily activity. Re-running the same source entry must not create another task, but work logged on different dates is new activity even when the subject, activity type, milestone, and scheduled-end month are identical.
• Milestone-first. Every CRM activity must link to a milestone via the regardingobjectid field. If no milestone exists, queue the activity locally and alert the user — never create an orphaned task.
• User reviews before CRM writes. In interactive mode, always present the proposed activities and get confirmation before creating anything. In automated mode, create activities and send a receipt summary.
• Preserve source language. Activity subjects and descriptions stay in their original language (English or Hebrew). Do not translate.
• Portable by design. No hardcoded user names, paths, or customer mappings. Everything is discovered at runtime via M365 profile + CRM queries + local folder scanning. Any team member with the prerequisites can run this skill.

Prerequisite Auto-Install

Before running, verify all dependencies are present. Install anything missing automatically.

Required Sibling Skills

This skill requires sibling skills from the same repository (https://github.com/roie9876/clawpilot-skills):

Check if each required skill is installed:

# macOS / Linux
for skill in daily-activity-log customer-repo msx-crm; do
  [ -f "$HOME/.copilot/skills/$skill/SKILL.md" ] && echo "✅ $skill" || echo "❌ $skill MISSING"
done

# Windows
foreach ($skill in @('daily-activity-log','customer-repo','msx-crm')) {
    if (Test-Path "$HOME\.copilot\skills\$skill\SKILL.md") { "✅ $skill" } else { "❌ $skill MISSING" }
}

If ANY required skill is missing, install all skills from the repository:

1. Clone the repo (skip if already cloned):

   # macOS / Linux
   [ -d "$HOME/customer-skills/.git" ] || git clone https://github.com/roie9876/clawpilot-skills.git "$HOME/customer-skills"
   

   # Windows
   if (-not (Test-Path "$HOME\customer-skills\.git")) {
       git clone https://github.com/roie9876/clawpilot-skills.git "$HOME\customer-skills"
   }
   

2. Run the installer (idempotent — safe to re-run):

   # macOS / Linux
   bash "$HOME/customer-skills/scripts/install.sh"
   

   # Windows
   pwsh "$HOME\customer-skills\scripts\install.ps1"
   

3. Verify the required skills are now installed. If still missing, stop and report the error.

Required Tools

Install any missing tools before proceeding. After installing Node.js, verify: node --version.

CRM Tool Script (run-tool.mjs)

The CRM helper script must exist at $HOME/.copilot/skills/msx-crm/crm-tools/run-tool.mjs.

# macOS / Linux
[ -f "$HOME/.copilot/skills/msx-crm/crm-tools/run-tool.mjs" ] && echo "✅ CRM tools found" || echo "❌ CRM tools missing"

# Windows
if (Test-Path "$HOME\.copilot\skills\msx-crm\crm-tools\run-tool.mjs") { "✅ CRM tools found" } else { "❌ CRM tools missing" }

If missing, clone the SE Kanban Tracker repo:

# macOS / Linux
git clone https://github.com/roie9876/crm-tools.git "$HOME/.copilot/skills/msx-crm/crm-tools"
cd "$HOME/.copilot/skills/msx-crm/crm-tools"

# Windows
git clone https://github.com/roie9876/crm-tools.git "$HOME\.copilot\skills\msx-crm\crm-tools"
Set-Location "$HOME\.copilot\skills\msx-crm\crm-tools"; npm install

VPN Connection

CRM is only accessible on Microsoft corpnet.

macOS / Linux: Run $HOME/Scripts/ensure-vpn.sh if it exists. If not, ask user to connect Azure VPN manually.

Windows: No automated VPN script. Ask user to confirm Azure VPN Client is connected.

M365 Sign-In

Check m_m365_status. If not signed in → call m_m365_sign_in.

Additional Checks

Upstream dependency: /daily-activity-log must run before /crm-activity-sync to populate activity-log.md. If activity-log.md is missing or has no entry for the target date, this skill reports "No activities logged for {date}. Run /daily-activity-log first."

Commands

File Locations

CRM Tool Path

All CRM operations use the existing tool runner:

~/Scripts/ensure-vpn.sh --quiet
node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs <tool-name> '<json-params>'

Important: The CRM tool currently supports read-only operations. Before this skill can create activities, run-tool.mjs must be extended with a create_task tool. See Phase 0: CRM Write Tool below.

Step 0: Environment Check

Run before every sync operation.

1. VPN: Run ~/Scripts/ensure-vpn.sh --quiet. If exit code ≠ 0:

   • Log: "⚠️ VPN not connected. CRM operations will fail."
   • If interactive → ask user to connect manually.
   • If automated → queue all activities to pending.json, send Teams notification: "CRM activity sync skipped — VPN down."
   • Stop.

2. CRM auth: Run node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs crm_auth_status.

   • If authenticated: false → same behavior as VPN failure.

3. Config: Check ~/.copilot/crm-activity-sync/config.json exists.

   • If missing → run Step 1 (setup) automatically.

4. Activity logs exist: Check that at least one activity-log.md file exists across projects.

   • If none exist → prompt: "No activity logs found. Run /daily-activity-log first to populate them."
   • In automated mode → send Teams notification and stop.

Step 1: First-Run Setup (/crm-activity-sync setup)

This step runs once to bootstrap the configuration. Re-run anytime to update.

1a. Discover User Identity

m365_get_my_profile → displayName, mail, jobTitle, department, officeLocation
m365_get_my_manager → manager name, email, title
CRM: crm_whoami → UserId (CRM system user GUID)

Store in config:

{
  "user": {
    "name": "Roie Ben Haim",
    "email": "robenhai@microsoft.com",
    "initials": "RBH",
    "jobTitle": "Azure Specialist SE",
    "department": "IL-Spec Sales",
    "officeLocation": "HERZLIYA-AT3",
    "crmUserId": "<GUID from WhoAmI>",
    "manager": {
      "name": "Ido Katz",
      "email": "idokatz@microsoft.com"
    }
  }
}

Initials derivation: Take the first letter of each word in the display name, uppercase. "Roie Ben Haim" → "RBH". Confirm with user — they may prefer different initials (e.g., "RB" instead of "RBH").

1b. Discover Customers and Projects

Scan the local customer-engagements folder:

ls ~/customer-engagements/

For each customer folder found:

1. Read ~/customer-engagements/{customer}/README.md — extract customer display name.
2. Read ~/customer-engagements/{customer}/stakeholders.md — extract Microsoft account team members (SSPs, account leads).
3. List ~/customer-engagements/{customer}/projects/ — each subfolder is a project.
4. For each project, read projects/{project}/README.md — extract project type (POC, ADS, etc.) and status.

1b-2. Classify Customers: Core vs. Temp

During setup, ask the user:

1. "Who are your dedicated SSPs?" — These are the field sellers you work with daily.
2. "Which customers are you actively covering?" — These become core customers.

Any customer folder in ~/customer-engagements/ that is NOT listed as core is classified as temp.

Core customers:

• Always scanned for activities during daily sync.
• Their SSPs are watched for new opportunities (/crm-activity-sync watch).
• Persist indefinitely in the config.

Temp customers:

• Scanned for activities the same way as core (customer-engagements is always source of truth).
• Their SSPs are NOT watched for new opportunities.
• After preferences.temp_archive_days (default 14) with no new activity, the sync prompts: "No activity with {customer} for {N} days. Archive this temp engagement?"
• Archiving sets customer_type: "archived" in crm-mapping.json — stops scanning but preserves history.

Classification is stored in crm-mapping.json:

{
  "customer_type": "core",  // or "temp" or "archived"
  "temp_reason": "Urgent help request via Ido Katz",  // only for temp
  "temp_started": "2026-04-20"  // only for temp
}

Discovery of new temp customers: When /daily-activity-log detects a calendar meeting with an external domain not matching any existing crm-mapping.json, it alerts:

• "Meeting detected with unknown customer domain: clal-ins.co.il. Is this: A) A new core customer, B) A temp engagement, C) Not customer work (skip)"

1c. Match Customers to MSX Accounts + Opportunities

For each discovered customer:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs list_opportunities '{"customerKeyword":"<customer-name>"}'

Present the matches to the user:

Found for "checkpoint":
  1. Checkpoint - SASE (opportunityid: xxx, owner: Kobi Shitrit)
  2. Checkpoint - CloudGuard (opportunityid: yyy, owner: ...)
Which opportunities are you actively working on?

For each confirmed opportunity, fetch milestones:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs get_milestones '{"opportunityId":"<opp-id>","statusFilter":"active"}'

1d. Build Per-Customer Mapping

Write ~/customer-engagements/{customer}/crm-mapping.json:

{
  "customer_display_name": "CheckPoint",
  "customer_slug": "checkpoint",
  "account_name": "CHECK POINT SOFTWARE TECHNOLOGIES LTD",
  "account_id": "<GUID>",
  "ssps": [
    {
      "name": "Kobi Shitrit",
      "email": "kobishitrit@microsoft.com",
      "role": "Account Lead",
      "crm_user_id": "<GUID>"
    }
  ],
  "projects": {
    "sase": {
      "display_name": "SASE",
      "project_type": "POC",
      "opportunity_id": "<GUID>",
      "opportunity_name": "Checkpoint - SASE",
      "milestones": [
        {
          "milestone_id": "<GUID>",
          "milestone_number": "7-502749008",
          "milestone_name": "SASE POC",
          "milestone_status": "On Track",
          "is_primary": true
        }
      ],
      "default_activity_type": "POC",
      "subject_keywords": ["sase", "checkpoint", "aks", "dpdk"],
      "work_repos": [
        {
          "path": "/Users/robenhai/SASE",
          "label": "SASE Design & POC",
          "added": "2026-04-23"
        }
      ]
    }
  },
  "last_refreshed": "2026-04-23T08:00:00Z"
}

Mapping rules:

• Each project maps to exactly one opportunity (1:1).
• A project may have multiple milestones (e.g., "SASE ADS" and "SASE POC"). Mark one as is_primary — this is the default target for activities.
• subject_keywords are used by /daily-activity-log to match calendar events to projects.
• default_activity_type is the fallback when classification is ambiguous.
• work_repos are registered manually — not every project has a repo. Used by /daily-activity-log to scan git history.

1e. Discover SSPs

For each confirmed opportunity:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs crm_query '{
  "entitySet": "msp_dealteams",
  "filter": "_msp_opportunityid_value eq '''<opportunity-id>'''",
  "select": "_msp_userid_value,msp_role",
  "top": 50
}'

Cross-reference with stakeholders.md to identify SSPs (Solution Sales Professionals), ATU members, and other field roles. Store in the ssps array of crm-mapping.json.

Also store the aggregated SSP list in the global config for opportunity watching:

{
  "known_ssps": [
    {
      "name": "Kobi Shitrit",
      "email": "kobishitrit@microsoft.com",
      "crm_user_id": "<GUID>",
      "customers": ["checkpoint"]
    }
  ]
}

1f. Set Preferences

Ask the user:

1. Sync mode: Interactive (review before creating) or Auto (create + receipt)?
   • Default: Interactive.
2. Scheduling: Manual only, or daily automation (Sun–Thu)?
   • If daily: what time? Default: 08:00 IST.
3. Activity naming convention:
   • Default: {initials} - {type} (e.g., "RBH - POC"). Matches the pattern in the screenshot.
   • Option: {initials} - {type} - {description} (e.g., "RBH - POC - AKS Architecture Review").
4. Which data sources to scan: Calendar (always on), customer-engagements (always on), Teams (optional), Kanban (optional).

Store preferences in config:

{
  "preferences": {
    "sync_mode": "interactive",
    "schedule": "daily",
    "schedule_time": "08:00",
    "schedule_days": ["sunday", "monday", "tuesday", "wednesday", "thursday"],
    "subject_format": "{initials} - {type}",
    "data_sources": ["calendar", "customer-engagements", "teams"],
    "auto_retry_pending": true
  }
}

1g. Confirm and Save

Present a summary of the entire config to the user:

✅ Setup complete!

User: Roie Ben Haim (RBH), Azure Specialist SE
Manager: Ido Katz
CRM User ID: xxx

Customers:
  • CheckPoint → Checkpoint - SASE → Milestone: SASE POC (7-502749008)
  • Clal Insurance → [opportunity TBD]
  • Sapiens → [opportunity TBD]

SSPs watching: Kobi Shitrit (checkpoint)
Sync mode: Interactive
Schedule: Daily Sun–Thu at 08:00

Ready to sync? Run /crm-activity-sync

Step 2: Read Activity Logs

This is the core data-reading phase. For a target date (or date range), scan all customer projects for logged activities.

Default target: Yesterday, adjusted for Israeli work week.

• If today is Sunday → yesterday was Saturday (weekend) → use Thursday.
• If today is Friday/Saturday → use Thursday.
• Otherwise → use yesterday.

2a. Scan All Projects

find ~/customer-engagements/*/projects/*/activity-log.md -type f 2>/dev/null

For each activity-log.md found:

1. Parse the file for ## <target-date> headers.
2. If a matching date header exists → extract the full entry (everything between this ## header and the next ## header or --- separator).
3. If no matching date → skip this project.

2b. Parse Each Entry

From each daily entry, extract:

2c. Build Activity List

Produce a list of activities to sync:

[
  {
    customer: "checkpoint",
    project: "sase",
    date: "2026-04-21",
    activity_type: "Design",
    summary: "Full design sprint for SASE — 10 architecture areas...",
    sources: "repo (21 commits)",
    already_synced: false  // checked against sync-log.json
  },
  ...
]

If zero activities found across all projects → report: "No activities logged for {date}. Run /daily-activity-log first." Stop.

Step 3: Map Activities to Milestones

For each activity in the list:

1. Load CRM mapping — read ~/customer-engagements/{customer}/crm-mapping.json.

   • If missing → queue with reason no_crm_mapping.

2. Find the project — look up projects.{project} in the mapping.

   • If missing → queue with reason project_not_mapped.

3. Select milestone — use the is_primary: true milestone for the project.

   • If the project has multiple milestones, check if the activity type matches a milestone name more closely (e.g., "Workshop" activity + milestone named "SASE Workshop" → use that one).

4. Validate milestone is still active — query CRM to confirm:

   node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs get_milestones '{"milestoneId":"<milestone-id>"}'
   

   Check msp_milestonestatus is in the active set (Not Started, On Track, In Progress, Blocked, At Risk).

5. Handle missing/closed milestones:

   Scenario	Action
   Milestone exists and is active	✅ Proceed to Step 4
   Milestone exists but is completed/closed	⚠️ Alert: "Milestone {name} is closed. Create activity anyway, or skip?"
   No milestone for this project	🔴 Queue to pending.json with reason no_milestone. Alert: "No active milestone for {customer}/{project}. Ask {ssp} to create one?"
   No opportunity for this customer	🔴 Queue with reason no_opportunity. Alert: "No MSX opportunity found for {customer}."

Step 4: Create CRM Activities

5a. Prepare Activity Payload

For each activity to create, build the CRM task record:

{
  "subject": "RBH - POC",
  "description": "SASE architecture review with CheckPoint — discussed AKS networking gaps, SRv6 encapsulation options, and multi-NIC requirements.",
  "scheduledend": "2026-04-30T08:00:00Z",
  "msp_taskcategory": 861980005,
  "regardingobjectid_msp_engagementmilestone@odata.bind": "/msp_engagementmilestones(<milestone-id>)"
}

Field mapping:

msp_taskcategory Mapping — All 20 Categories:

The **Type:** line in activity-log.md uses the exact CRM label. This skill maps it directly to the msp_taskcategory integer value:

If the Type: value from activity-log.md doesn't match any CRM label exactly, fall back to the project's default_activity_type from crm-mapping.json. If that's also missing, default to Customer Engagement (861980000).

Subject format options (from config subject_format):

• {initials} - {type} → "RBH - POC" (matches screenshot pattern)
• {initials} - {type} - {description} → "RBH - POC - AKS Architecture Review"
• {initials} - {type} ({date}) → "RBH - POC (Apr 22)"

Subject {type} short labels (for readability in CRM):

5b. Idempotency Check

Build a stable source key for each parsed daily entry:

{activity-date}|{customer}|{project}|{subject}|{milestone-id}

Check sync-log.json for that exact key before creating the task. Also include Activity date: YYYY-MM-DD in the CRM task description so the source date is preserved and can be used for recovery if the local sync log is unavailable.

A task is a duplicate only when it represents the same source activity date, customer, project, subject, and milestone. A same-subject task from another day or another month is not a duplicate and must not be skipped. For example, daily RBH - POC entries on the same milestone are separate CRM activities when their activity dates differ.

Before creating, optionally query milestone activities for recovery:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs get_milestone_activities '{"milestoneId":"<milestone-id>"}'

Only treat a CRM result as a recovered duplicate when its subject matches and its description contains the same Activity date: YYYY-MM-DD marker. Never deduplicate by subject or scheduled-end month alone.

If the exact source key or date marker matches:

• Interactive: "This exact daily activity is already synced. Skip or update?"
• Auto: Skip and retain the existing task ID in sync-log.json.

5c. Ensure Deal Team Membership

Before creating any task on a milestone, verify the user is on the deal team for the parent opportunity. If not, auto-add them.

Check membership:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs crm_query '{
  "entitySet": "msp_dealteams",
  "filter": "_msp_parentopportunityid_value eq '"'"'<opportunity-id>'"'"' and _msp_dealteamuserid_value eq '"'"'<user-crm-id>'"'"' and statecode eq 0",
  "select": "msp_dealteamid",
  "top": 1
}'

• If count > 0 → already on team, proceed to task creation.
• If count == 0 → auto-add to deal team:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs crm_query '{
  "method": "POST",
  "entitySet": "msp_dealteams",
  "body": {
    "msp_dealteamuserid@odata.bind": "/systemusers(<user-crm-id>)",
    "msp_parentopportunityid@odata.bind": "/opportunities(<opportunity-id>)"
  }
}'

On success:

• Log: "Added {user} to deal team for {opportunity-name}."
• Cache in crm-mapping.json: "deal_team_verified": true on the project.
• Proceed to task creation.

On failure (403/permissions):

• Interactive: "Cannot auto-join deal team for {opp}. Ask SSP ({ssp-name}) to add you."
• Auto: send Teams message to user with the SSP name. Queue activity in pending.json.

Cache the check. Once verified, set "deal_team_verified": true in the project's crm-mapping.json entry. Skip the check on subsequent syncs unless the opportunity changes. Reset to false if the opportunity ID changes or a task creation fails with a permissions error.

5d. Create the Task

This requires the create_task tool in run-tool.mjs (see Phase 0).

Parse duration from activity-log.md: Extract the number from the **Duration:** line (e.g., **Duration:** 285 min (4h 45m) — ... → 285). Pass as duration (the tool maps it to actualdurationminutes internally).

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs create_task '{
  "subject": "RBH - POC",
  "description": "...",
  "scheduledend": "2026-04-30T08:00:00Z",
  "milestoneId": "<milestone-guid>",
  "taskcategory": 861980005,
  "duration": 285
}'

Duration field mapping:

• duration — maps to actualdurationminutes CRM field internally
• Value in minutes, rounded to nearest 15 by /daily-activity-log
• If **Duration:** line is missing from activity-log.md (older entries), omit the field — CRM will use its default

On success:

• Log to sync-log.json with the created task ID.
• In interactive mode: "✅ Created 'RBH - POC' on milestone SASE POC."
• In auto mode: append to receipt.

On failure:

• If permissions error → deal team membership may have been revoked. Reset deal_team_verified to false, retry Step 5c, then retry task creation.
• Save to pending.json with error details.
• In interactive mode: show error and ask to retry.
• In auto mode: send Teams notification.

5d. Present Summary

After all activities are processed, show a summary:

📋 CRM Activity Sync — April 22, 2026

Created:
  ✅ RBH - POC → SASE POC (CheckPoint) — AKS architecture review
  ✅ RBH - Design → AI Search ADS (Clal Insurance) — Agent infrastructure discussion

Skipped (already exists):
  ⏭️ RBH - Meeting Call → SASE POC (CheckPoint) — duplicate

Pending (no milestone):
  🔴 Sapiens / Digital-AI — no active milestone. Queued.

Run /crm-activity-sync retry after milestone is created.

Step 5: Milestone Scanner and Opportunity Watch

This step runs as part of every sync. It serves three purposes:

1. Find new opportunities from core SSPs where you're not on the deal team.
2. Find new milestones on tracked opportunities.
3. Resolve orphaned activities from the pending queue.

5a. Scan Core SSP Opportunities

For each SSP in config.json → core_ssps:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs crm_query '{"entitySet":"opportunities","filter":"_ownerid_value eq '"'"'<ssp-crm-user-id>'"'"' and statecode eq 0","select":"opportunityid,name,_parentaccountid_value,createdon","orderby":"createdon desc","top":20}'

Compare against known opportunities in all crm-mapping.json files. If an opportunity is:

• Owned by a core SSP
• For an account that matches a customer you cover
• NOT already in any crm-mapping.json
• Created in the last 30 days (or new since last check)

Then alert:

🔔 New opportunity detected!
  "Checkpoint - CloudGuard" by Kobi Shitrit
  → Add to your tracking? (Yes / Skip)

5b. Refresh Milestones on Tracked Opportunities

For each opportunity in all crm-mapping.json files:

node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs get_milestones '{"opportunityId":"<opp-id>","statusFilter":"active"}'

Compare against milestones currently in the mapping:

• New milestone found → "New milestone 'SASE Workshop' on Checkpoint - SASE. Add to mapping?"
• Milestone status changed (e.g., On Track → Completed) → update the mapping.
• Milestone that was missing now exists → check if any pending activity matches.

5c. Resolve Orphaned Activities

For each pending activity with reason no_milestone:

1. Check mapped opportunity — query milestones again (SSP may have created one since last check).

2. If still no milestone on mapped opp — broaden search to ALL active opportunities for the customer account:

   node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs list_opportunities '{"customerKeyword":"<customer-name>"}'
   

   Then get milestones for each:

   node ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs get_milestones '{"opportunityId":"<opp-id>","statusFilter":"active"}'
   

3. Score each discovered milestone for the pending activity:

   • Milestone name contains project keyword? +3
   • Milestone category matches activity type (e.g., POC → POC/Pilot category)? +2
   • Milestone owner is a known SSP? +1
   • Milestone date is in the future? +1

4. If best score ≥ 3 → suggest: "Your pending {customer}/{project} activity might match milestone '{name}' on '{opportunity}'. Map it?"

5. If best score < 3 → no good match. In interactive mode: "No milestone found for {customer}/{project}. Notify SSP ({ssp_name}) to create one?"

6. If resolved → update crm-mapping.json with the new milestone, create the CRM task, move from pending to sync-log.

5d. When Does the Scanner Run?

Step 6: Pending Queue Management

pending.json Structure

[
  {
    "id": "pending-001",
    "customer": "sapiens",
    "project": "digital-ai",
    "date": "2026-04-22",
    "activity_type": "Meeting Call",
    "subject": "RBH - Meeting Call",
    "description": "Digital AI overview session",
    "reason": "no_milestone",
    "evidence": {
      "calendar_event_id": "AAMkAGI2...",
      "meeting_note": "~/customer-engagements/sapiens/projects/digital-ai/meetings/2026-04-22-digital-ai-overview.md"
    },
    "created_at": "2026-04-23T08:00:00Z",
    "retry_count": 0,
    "last_retry": null
  }
]

Retry Logic (/crm-activity-sync retry)

1. Load pending.json.
2. For each pending item: a. Re-check the reason:
   • no_milestone → query milestones again. If one now exists → proceed to create.
   • no_opportunity → query opportunities again.
   • crm_error → retry the create operation.
   • vpn_down / auth_failed → check VPN/auth, retry if resolved. b. If resolved → create the task, move from pending to sync-log. c. If still blocked → increment retry_count, update last_retry.
3. Report results.

Auto-Retry

If preferences.auto_retry_pending is true, every daily sync run starts by retrying pending items before processing new activities.

Stale Pending Cleanup

Items pending for >30 days are flagged:

⚠️ Stale pending items (>30 days):
  - Sapiens / Digital-AI (Apr 22) — no_milestone — 31 days pending
  Archive or escalate?

Failure Scenarios

Phase 0: CRM Write Tool

This must be implemented before the skill can create activities.

Add create_task to ~/.copilot/skills/msx-crm/crm-tools/run-tool.mjs:

async create_task({ subject, description, scheduledend, milestoneId, taskcategory, ownerId }) {
  if (!subject) error('subject is required');
  if (!milestoneId) error('milestoneId is required');
  const nid = normalizeGuid(milestoneId);
  if (!isValidGuid(nid)) error('Invalid milestoneId GUID');

  const body = {
    subject,
    ...(description && { description }),
    ...(scheduledend && { scheduledend }),
    ...(taskcategory && { msp_taskcategory: taskcategory }),
    'regardingobjectid_msp_engagementmilestone@odata.bind':
      `/msp_engagementmilestones(${nid})`
  };

  // If ownerId provided, bind it; otherwise CRM defaults to current user
  if (ownerId) {
    const oid = normalizeGuid(ownerId);
    if (isValidGuid(oid)) {
      body['ownerid@odata.bind'] = `/systemusers(${oid})`;
    }
  }

  const result = await crm.request('tasks', {
    method: 'POST',
    body
  });

  if (!result.ok) {
    error(`Create task failed (${result.status}): ${result.data?.message}`);
  }

  // Extract the created task ID from the OData-EntityId header or response
  const taskId = result.data?.activityid || 'created (ID not returned)';
  return { success: true, taskId, subject };
}

Also add update_task for future use (updating description, closing tasks):

async update_task({ taskId, subject, description, scheduledend, statuscode, statecode }) {
  if (!taskId) error('taskId is required');
  const nid = normalizeGuid(taskId);
  if (!isValidGuid(nid)) error('Invalid taskId GUID');

  const body = {};
  if (subject !== undefined) body.subject = subject;
  if (description !== undefined) body.description = description;
  if (scheduledend !== undefined) body.scheduledend = scheduledend;
  if (statuscode !== undefined) body.statuscode = statuscode;
  if (statecode !== undefined) body.statecode = statecode;

  if (!Object.keys(body).length) error('No fields to update');

  const result = await crm.request(`tasks(${nid})`, {
    method: 'PATCH',
    body
  });

  if (!result.ok) {
    error(`Update task failed (${result.status}): ${result.data?.message}`);
  }

  return { success: true, taskId: nid };
}

Testing the write tools:

1. First test with crm_whoami to confirm auth.
2. Create a test task on a known milestone with a clearly-marked test subject.
3. Verify it appears in MSX UI.
4. Delete or close the test task.

Automation Setup

If the user opts for daily scheduling during setup, create two Clawpilot automations:

Automation 1: Daily Activity Log
Schedule: Every weekday at 07:30 (Sun-Thu Israel time)
Prompt: |
  Run /daily-activity-log for yester

…(truncated)