feat(rmm-audit): add roadmap reconciliation pass (Agent F) + cleanup step

The skill's frontmatter claimed it updated FEATURE_ROADMAP.md, but the body
had no roadmap-reconciliation logic — so stale checkboxes slipped through
(Network Discovery Node backend + BUG-001 temperature both shipped while
marked [ ]). Added:
- Agent F (parallel, read-only): cross-references every roadmap checkbox
  against code artifacts; classifies STALE-INCOMPLETE / PARTIAL /
  STALE-COMPLETE / ACCURATE with proving artifact. Conservative — only
  flips when end-to-end evidence is unambiguous; backend/scaffolding-only
  is PARTIAL, never flipped.
- Living-docs step: actually flip stale checkboxes, annotate partials,
  flag [x]-but-missing as [HIGH] regressions; every change logged in the
  report's new "FEATURE_ROADMAP.md Delta" section (no silent edits).
- Phase 0 extracts the roadmap claims list; --pass=roadmap added.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.claude/skills/rmm-audit/SKILL.md

@@ -10,7 +10,8 @@ description: |
   FEATURE_ROADMAP.md). Takes 10-20 minutes.
 
   Invoke explicitly only — no auto-trigger. Use /rmm-audit for a full audit.
-  Optional arg: --pass=<name> to run a single pass (api, ui, rust, ts, security, pipeline).
+  Optional arg: --pass=<name> to run a single pass (api, ui, rust, ts, security, pipeline, roadmap).
+  The roadmap pass reconciles FEATURE_ROADMAP.md checkboxes against the code and cleans up stale ones.
 ---
 
 # GuruRMM End-to-End Audit
@@ -58,12 +59,17 @@ Before spawning agents, read these yourself:
 Capture from `server/src/api/mod.rs` the complete route list (all `.route(...)` calls).
 This becomes the **authority route list** passed to both the API and UI audit agents.
 
+Also extract every checkbox line from `FEATURE_ROADMAP.md` (both `[ ]` and `[x]`, with
+section + priority) into a **roadmap claims list** — passed to Agent F for reconciliation
+against the code.
+
 ---
 
 ## Phase 1: Parallel Audit Agents
 
-Spawn all four agents simultaneously in a single message. Each agent receives the
+Spawn all five parallel agents (A, B, C, D, F) simultaneously in a single message.
-full context it needs inline — do not assume they share context.
+Each agent receives the full context it needs inline — do not assume they share context.
+(Agent E — Build Pipeline — runs sequentially afterward; it SSHes the live build server.)
 
 ---
 
@@ -226,6 +232,48 @@ Return structured findings with file:line references.
 
 ---
 
+### Agent F — Roadmap Reconciliation
+
+**Goal:** Make `FEATURE_ROADMAP.md` tell the truth. The roadmap drifts from reality:
+features ship in code while their checkbox stays `[ ]` (verified cases: the Network
+Discovery Node backend, and temperature monitoring / BUG-001 — both shipped while marked
+incomplete), and occasionally an item is `[x]` but the code was never written or was
+reverted. This pass reconciles every checkbox against the code.
+
+**Runs in parallel with A–D (read-only code search, no SSH).**
+
+**Instructions for agent:**
+
+1. Read `docs/FEATURE_ROADMAP.md` and extract every checkbox line with its state
+   (`[ ]` / `[x]`), section, and priority.
+
+2. For EACH item, find the implementing artifact in the code — do NOT trust the
+   checkbox. Map feature → evidence:
+   - API/endpoint features → `server/src/api/mod.rs` routes + the handler module
+   - DB/schema features → `server/migrations/*.sql` + `server/src/db/*.rs`
+   - Agent capabilities → `agent/src/` modules
+   - Dashboard/UI features → `dashboard/src/pages|components/*.tsx` + `api/client.ts`
+   Use Grep/GrepAI. Cite the exact artifact (file:line, migration name, route path).
+
+3. Classify each item:
+   - **STALE-INCOMPLETE** — `[ ]` but the code fully implements it end-to-end →
+     recommend flip to `[x]`. Cite the proving artifact.
+   - **PARTIAL** — `[ ]`, backend/some layers exist but not end-to-end for the stated
+     scope (e.g. backend + API done but no UI; scaffolding present but unwired — like
+     safe-rollout, or discovery with a per-agent tab but no fleet view). Keep `[ ]` but
+     recommend an inline annotation, e.g. `[ ] ... (backend done; UI/fleet-view pending)`.
+     Do NOT flip partials to `[x]`.
+   - **STALE-COMPLETE** — `[x]` but the code does not implement it (never built or
+     reverted) → recommend flip to `[ ]`, flag `[HIGH]` (the roadmap is lying).
+   - **ACCURATE** — checkbox matches code. No change.
+
+4. Be conservative: only flip `[ ]`→`[x]` when the evidence is unambiguous AND the
+   feature is complete end-to-end for its stated scope. When in doubt → PARTIAL with a
+   note, never a flip. A backend-only or scaffolding-only implementation is PARTIAL.
+
+Return a table:
+`item text | section | current state | verdict (STALE-INCOMPLETE / PARTIAL / STALE-COMPLETE / ACCURATE) | proving-or-missing artifact`
+
 ---
 
 ### Agent E — Build Pipeline Health
@@ -375,7 +423,7 @@ Write to: `projects/msp-tools/guru-rmm/reports/YYYY-MM-DD-rmm-audit.md`
 # GuruRMM Audit Report — YYYY-MM-DD
 
 **Auditor:** Claude (claude-opus-4-7)
-**Passes:** API Coverage, UI Gaps, Rust Quality, TypeScript Quality, Data Integrity, Build Pipeline
+**Passes:** API Coverage, UI Gaps, Rust Quality, TypeScript Quality, Data Integrity, Build Pipeline, Roadmap Reconciliation
 **Previous audit:** [link to prior report if one exists, else "First audit"]
 
 ---
@@ -452,6 +500,19 @@ New gaps discovered this audit:
 
 ---
 
+## FEATURE_ROADMAP.md Delta (Agent F)
+
+Checkboxes corrected to match code reality this audit:
+- `[ ]` → `[x]` **<item>** — was marked incomplete but is shipped. Proof: `<artifact>`.
+- `[x]` → `[ ]` **<item>** — marked complete but NOT in code (regression / never built). `[HIGH]`
+
+Annotated as partial (left `[ ]`, scope clarified):
+- `[ ]` **<item>** (backend done; UI/fleet-view pending) — `<artifact>`
+
+Verified accurate (no change): N items.
+
+---
+
 ## Recommended Action Order
 
 1. [CRITICAL items, sorted by impact]
@@ -467,6 +528,24 @@ After writing the report, update `docs/UI_GAPS.md`:
 - Add any newly discovered gaps under the appropriate priority section
 - Do NOT remove completed items — move them to the "Completed Features" section
 
+### Update FEATURE_ROADMAP.md (roadmap cleanup)
+
+Apply Agent F's reconciliation to `docs/FEATURE_ROADMAP.md` — this is the roadmap
+cleanup the audit is responsible for. The roadmap is a living doc, so editing it fits
+the "living docs are updated" exception to read-only.
+
+- **STALE-INCOMPLETE → flip `[ ]` to `[x]`** for every item Agent F proved is shipped
+  end-to-end. Keep the line text; optionally append `(verified <date>)`.
+- **PARTIAL → leave `[ ]`, append the scope annotation** Agent F recommends
+  (e.g. `(backend done; UI/fleet-view pending)`). Never flip a partial to `[x]`.
+- **STALE-COMPLETE → flip `[x]` to `[ ]`** and add `[REGRESSION — flagged YYYY-MM-DD]`
+  so it's visible; mirror it as a `[HIGH]` finding in the report.
+- Update any "Last Updated" / status line in the roadmap.
+- Every checkbox change MUST also appear in the report's "FEATURE_ROADMAP.md Delta"
+  section with its proving (or missing) artifact — no silent edits.
+- When evidence is ambiguous, do NOT change the checkbox; record it as PARTIAL/INFO
+  and leave the cleanup for human review. Bias toward under-flipping.
+
 ---
 
 ## Phase 5: User Summary
@@ -482,6 +561,7 @@ MEDIUM (N):   Batched in report.
 
 Pipeline: [one-line status — e.g. "all green" or highest-severity finding]
 UI_GAPS.md: N items marked complete, N new gaps added.
+Roadmap: N checkboxes corrected (N stale-incomplete flipped to done, N partial annotated, N regressions flagged).
 
 Recommended first action: [the single highest-priority finding]
 ```