====================================================================
 FNMA AI LENDER LETTER KIT — START HERE
====================================================================

This kit runs as a SKILL inside one Claude PROJECT. The Skill carries
the rules, the verbatim rule kit, and the output format — so you only
ever upload your own P&Ps. Install it once and never paste a prompt or
attach the CSV again.

You need Claude Pro / Max / Team / Enterprise (the Skill needs Code
execution). Do all the work inside one Project.

THE FLOW:
  inventory each batch -> reconcile (add your vendor list etc.)
  -> one synthesis chat -> realign each batch -> change log.

Each output is a working draft for compliance/attorney review —
NOT legal advice.

(test-findings.md is the output of a local test — reference only, you
don't need it. _skill-src/ is the unpacked Skill, for reading; you
install the zip, not the folder.)


--------------------------------------------------------------------
 STEP 0 — INSTALL THE SKILL (once)
--------------------------------------------------------------------
  1. Settings -> Capabilities -> turn ON "Code execution."
  2. Settings -> Customize -> Skills -> "+" -> Upload a skill ->
     pick  fnma-ai-lender-letter-skill.zip  -> toggle it ON.
  3. Make a Project. (Do NOT put the zip in the project — it's a
     Skill, not a project file.)

New to installing Skills? The USER GUIDE folder has a
screenshot-by-screenshot walkthrough.


--------------------------------------------------------------------
 STEP 1 — INVENTORY EACH BATCH
--------------------------------------------------------------------
New chat -> upload one batch of P&Ps -> paste this whole box:

    Inventory this batch for the FNMA AI lender letter.

Repeat for every batch (new chat + upload + paste, each time) until
all batches are done.


--------------------------------------------------------------------
 STEP 2 — RECONCILE  (one chat, after ALL batches — fill the gaps)
--------------------------------------------------------------------
The P&P scan only sees what's written in the policies. This step adds
what they can't show — so FIRST drop your extra files into the project
(any you have): vendor inventory list, vendor contracts, IT/app
inventory, org chart. Then new chat -> paste this whole box:

    Consolidate the findings from all my inventoried batches into one
    working list. Then show me the gaps — including the AI the P&Ps
    can't show (embedded vendor AI, shadow/employee AI, tools added
    since the P&Ps were written). I'm giving you our vendor inventory,
    contracts, and app list in the project — cross-reference them,
    surface any system no P&P showed, and validate the owners. Work
    through the high-risk gaps with me, then save a reconciled
    working-inventory.md to the project folder.

Answer its questions as you go. When you're done, it saves
working-inventory.md — that's what the next step uses.


--------------------------------------------------------------------
 STEP 3 — SYNTHESIZE  (ONE chat)
--------------------------------------------------------------------
New chat -> paste this whole box:

    Build the master AI System Inventory, Risk Assessment, and AI
    Governance Policy from working-inventory.md in the project folder.
    List every open CONFIRM grouped by who I need to ask and what to
    ask. Save the master policy as a file in the project folder.

(Skipped Reconcile? Replace working-inventory.md with "every
findings-*.md file". On claude.ai web, use: Search this project's
chats for every ## FINDINGS block.)


--------------------------------------------------------------------
 STEP 4 — REALIGN THE P&Ps  (one batch at a time)
--------------------------------------------------------------------
In each batch's chat -> paste this whole box:

    Update these P&Ps to match the master AI policy in the project
    folder.

Do that for each batch.


--------------------------------------------------------------------
 STEP 5 — CHANGE LOG
--------------------------------------------------------------------
New chat -> paste this whole box:

    Make the Appendix-A change log for the master AI policy and every
    P&P we changed.


--------------------------------------------------------------------
That's the whole job: inventory each batch -> reconcile (add your
vendor list + work the gaps) -> ONE synthesis chat -> realign each
batch -> change log.

Notes: Synthesis is ONE chat after reconcile — not per batch.
Reconcile is where the vendor inventory / contracts / app list go in,
and where undocumented & shadow AI gets caught — don't skip it on a
real firm-wide run. On web, cross-chat search lags (paste FINDINGS
blocks if it misses some); on Cowork it reads the files, no lag.
Nothing to upload for the kit — it's bundled in the Skill; only add a
CSV if you're handed a newer one.
