/lynqu-lead-management

name: lynqu-lead-management
description: Keep the Lynqu pipeline moving. Stage moves, lead routing and owner assignment, scoring, tags, bulk hygiene and stalled deals. Requires the Lynqu MCP server connected.

Lynqu Pipeline Hygiene

You keep the board honest. A pipeline lies in predictable ways: deals sit in stages they left months ago, leads are assigned to people who no longer work here, the same person exists three times, and every forecast built on top of it is fiction.

You fix that: in bulk, safely, with the user's approval at each destructive step. The output is a board a manager can trust and a rep can work top-down.

Invocation

/lynqu pipeline [board | segment | "what's stalled" | "reassign Sara's leads"]

Step 1: Take the pipeline's temperature

• **`get-org-summary`**: the shape of the book
• **`list-lead-pipelines`** and **`get-pipeline-stages`**: the boards and their

columns. Everything downstream is per-board; there is no global stage list

• **`list-leads`** per board with stage and updated dates: the raw material
• **`list-lead-duplicates`**: the hygiene problem that corrupts every other number
• **`list-team-members`**: who's active, and who has left but still owns leads
• **`list-lead-views`**: the org's saved views already encode what it watches
• **`list-segments`**: the shared filters (market, source, campaign, tags, UTM)

with live counts and what each one drives: sequences, rules, A/B variants

Step 2: Diagnose before you touch anything

Six checks, in this order. Report all of them before fixing any of them: a manager needs the shape of the problem, not a running commentary of edits.

| Check | How to spot it | Why it matters | |-------|----------------|----------------| | **Stalled** | Days in current stage > 2× the board's median | The main source of forecast error | | **Untouched** | No activity and no completed contact point since creation | Leads that were captured and abandoned | | **Orphaned** | Assigned to a removed or suspended member | Nobody is working these, and nobody knows | | **Duplicates** | `list-lead-duplicates` | Splits history and double-counts value | | **Taskless** | Open lead, no open contact point | An open lead with no next step is a wish | | **Miscategorized** | Won/lost leads still sitting in an open stage | Silently inflates the pipeline |

Quantify each: *"41 stalled (median 9 days, these are 30+), 12 orphaned from Sara's departure, 6 duplicate pairs, 88 with no open task."*

Step 3: Decide on stalled deals, don't just report them

Every stalled lead gets one of three verdicts, and the user picks in bulk:

1. **Revive**: there's a real reason to re-engage → task + `lynqu-sales-followup` 2. **Nurture**: real fit, wrong timing → move to a nurture stage, set a dated task at the trigger, stop spending attention now 3. **Close as lost**, with a **reason**. This is the important one. A lost lead with a reason feeds `lynqu-icp`; a lead left open forever feeds nothing and corrupts the forecast. A closed lead also stops receiving automated follow-ups by itself

Nobody wants to close deals, which is exactly why the pipeline fills up. Make the recommendation explicit and let the human decide.

Step 4: Fix it

Show the full plan with counts, get approval, then work in this order, the sequence matters, because merging after reassigning means doing the work twice.

1. **Duplicates first.** `merge-leads`: pick the survivor deliberately (the one with the richer history and the correct owner) and show what reparents: notes, tasks, documents, activity, campaign links. A merge cannot be undone, so each pair gets its own confirmation, not a batch one. The same goes for duplicate company records splitting an account's roll-up: `merge-companies` with no arguments lists the pairs, then one confirmed call per pair 2. **Records that should never have existed** (test entries, spam submissions, a lead created by mistake): `delete-lead`. Never for a lost deal (close it, so it stays in reporting) and never for a duplicate (merge it, or half the history goes with the deleted copy). List every lead by name and id and get a yes for exactly that list. The lead leaves every board and report at once, and only an admin can restore it 3. **Ownership.** `assign-lead` for orphans and rebalancing. Takes the **user id**; `list-team-members` returns both that and the `organization_user_id` (which is what campaign membership and department assignment want) 4. **Stage moves.** `update-lead-stage` per the verdicts. `move-lead-pipeline` when a lead is on the wrong board entirely: a different board is a different process, not a different column 5. **Scores, tags, temperature.** `bulk-update-leads`, ≤ 100 per call. Leads scanned without the event picked: `attach-leads-to-event` (≤ 100), after saying that attaching runs the event's follow-up sequences, so it can send that event's first email 6. **Tasks.** `add-lead-contact-point` on everything that stays open, and `update-lead-contact-point` to close out what's already done. A board where every open lead has a dated next step is the definition of "working" 7. **Notes.** `add-lead-note` on anything whose state you changed for a non-obvious reason. Six weeks later, "why is this in nurture?" needs an answer. A note that is wrong gets corrected with `update-lead-note`; one that should not be there at all (pasted into the wrong lead, personal data that does not belong) is removed with `delete-lead-note`, after a yes that quotes it. Both work on the caller's own notes, or on others' for an admin or a manager over the author's department. Voice notes are transcripts and cannot be reworded

Step 5: Make it stick

A one-off cleanup decays in a month. Before you finish:

• Propose a **saved view** for the checks that mattered ("stalled > 30 days",

"no open task") so this becomes a filter, not a memory

• Propose a **s