pilotspace · pilotspacex-byte · Jul 10, 2026 · Jul 10, 2026 · Jul 10, 2026 · Jul 10, 2026
diff --git a/.add/GLOSSARY.md b/.add/GLOSSARY.md
@@ -1,5 +1,15 @@
 # GLOSSARY  (one name per concept — used everywhere: specs, contracts, code)
 
+## Sensitivity classes
+The risk-CLASS a task declares (`sensitivity:` in the TASK header) — read live by freeze/status/check.
+The base four (security · data · architecture · mechanical) are method-universal and always apply.
+
+Datetime, money, or timezone arithmetic in scope ⇒ declare `data` (the full lane): the value
+formats themselves are the risk surface — a suite testing a friendlier dialect than the spec's
+own examples stays green through a real crash (the benchmark wm2 aware/naive miss,
+results/2026-07-wv1-rep0.md).
+
+
 ADD: AI-Driven Development — the orchestration engine (the build/verify discipline) and this skill.
 AIDD: the umbrella method and the book that explains the why (the trust layer).
 Task: one feature taken through the flow; lives in `.add/tasks/<slug>/TASK.md`.

diff --git a/.add/SEAMS.md b/.add/SEAMS.md
@@ -54,7 +54,7 @@ Citations: 232 files reference "byte-identical" in `.add/tasks/` — method:
 
 ## scope-token-grammar
 Name: §5 "Scope (may touch):" token-resolution grammar
-Anchor: `add-method/tooling/add.py:5269` (`_declared_scope`)   <!-- re-pinned 2026-07-10 x16: 4786→4798→4870→4892→4896→4900→4918→5002→5011→5165→5173→5206→5219→5231→5269 (branch merge folded the remote tiny-plan-small-scope lane into cmd_new_task above it — the pin drifts on ANY upstream add.py change; symbol cited so the drift self-describes) -->
+Anchor: `add-method/tooling/add.py:5337` (`_declared_scope`)   <!-- re-pinned 2026-07-11 x18: 4786→…→5269→5324→5337 (fast-lane-boundary-line's cmd_freeze boundary guard above it — the pin drifts on ANY upstream add.py change; symbol cited so the drift self-describes) -->
 Contract: `_declared_scope` reads ONLY the first physical line after the §5 header — a
   wrapped multi-line list silently truncates. Each backticked token then resolves
   independently: `./...` = this task's dir, any token containing `/` = project-root-relative,

diff --git a/.add/milestones/quality-floors/RETRO.md b/.add/milestones/quality-floors/RETRO.md
@@ -0,0 +1,35 @@
+════════════════════════════════════════════════════════════════════════
+ quality-floors · Quality floors for lean ADD
+════════════════════════════════════════════════════════════════════════
+ VERDICT   DONE
+ TASKS     3/3 done           CRITERIA  0/0 met
+ GATES     3 PASS             WAIVERS   none
+
+ goal  convert the WV1 wm2 miss (own tests spoke a friendlier input
+       dialect than the spec's own examples → aware/naive datetime crash
+       shipped green) into narrow, ~zero-turn floors — the lean lane
+       keeps its −27–33% cost while the failure CLASS is closed; quality
+       is bought with mechanical floors, never with re-added ceremony
+ closed by Tin Dang <tindang.ht97@gmail.com>
+
+ TASK                        PHASE     GATE TESTS PROGRESS
+ ───────────────────────────────────────────────────────────────────────
+ spec-dialect-floor          done      PASS 6†    ●●●●●●●●●
+ dialect-check-and-data-voc… done      PASS 6†    ●●●●●●●●●
+ fast-lane-boundary-line     done      PASS 9†    ●●●●●●●●●
+ legend  ● reached  ◉ current  ○ pending   spec→…→done
+ † counted at the §4-declared path
+
+ GATED BY
+   spec-dialect-floor       PASS Tin Dang <tindang.ht97@gmail.com>
+   dialect-check-and-data-… PASS Tin Dang <tindang.ht97@gmail.com>
+   fast-lane-boundary-line  PASS Tin Dang <tindang.ht97@gmail.com>
+
+ EXIT CRITERIA  ○○○○○○○○○○ 0/0 met
+
+ LEARNINGS      none
+
+ SPEC DELTAS    27 open deltas — resolve: new-task --from-delta / drop-delta
+
+ DECIDE NEXT  consolidate learnings + archive-milestone quality-floors
+════════════════════════════════════════════════════════════════════════
diff --git a/.add/state.json b/.add/state.json
@@ -1,7 +1,7 @@
 {
   "project": "AIDD / ADD Methodology",
   "stage": "mvp",
-  "active_task": null,
+  "active_task": "persona-nudge-quiet",
   "tasks": {
     "add-check": {
       "title": "add.py check: validate .add project integrity",
@@ -7547,7 +7547,7 @@
     }
   },
   "created": "2026-05-28T15:39:04+00:00",
-  "updated": "2026-07-10T17:12:59+00:00",
+  "updated": "2026-07-10T22:50:47+00:00",
   "milestones": {
     "v1-1": {
       "title": "v1.1 \u2014 adoption & ergonomics",
@@ -7867,26 +7867,9 @@
         "email": "tindang.ht97@gmail.com",
         "source": "git"
       }
-    },
-    "quality-floors": {
-      "title": "Quality floors for lean ADD",
-      "goal": "",
-      "stage": "mvp",
-      "status": "active",
-      "created": "2026-07-10T17:12:27+00:00",
-      "updated": "2026-07-10T17:12:59+00:00",
-      "tiny": true,
-      "confirmed": true,
-      "confirmed_at": "2026-07-10T17:12:59+00:00",
-      "confirmed_by": "tindang",
-      "actor": {
-        "name": "Tin Dang",
-        "email": "tindang.ht97@gmail.com",
-        "source": "git"
-      }
     }
   },
-  "active_milestone": "quality-floors",
+  "active_milestone": "add-lean-loop",
   "archived": [
     {
       "slug": "v5",
@@ -8701,6 +8684,17 @@
         "skip-error-ergonomics"
       ],
       "archived": "2026-07-11"
+    },
+    {
+      "slug": "quality-floors",
+      "title": "Quality floors for lean ADD",
+      "tasks": 3,
+      "task_slugs": [
+        "spec-dialect-floor",
+        "dialect-check-and-data-vocab",
+        "fast-lane-boundary-line"
+      ],
+      "archived": "2026-07-11"
     }
   ],
   "active_milestones": [
@@ -8709,8 +8703,7 @@
     "build-strategy-facets",
     "delta-drain",
     "add-bench",
-    "add-lean-loop",
-    "quality-floors"
+    "add-lean-loop"
   ],
   "active_tasks": {
     "flow-simplification": "phase-review",

diff --git a/.add/tasks/dialect-check-and-data-vocab/TASK.md b/.add/tasks/dialect-check-and-data-vocab/TASK.md
@@ -0,0 +1,102 @@
+# TASK: §6 dialect check line + datetime/money/tz ⇒ data sensitivity vocabulary
+
+slug: dialect-check-and-data-vocab · created: 2026-07-11 · stage: mvp
+milestone: quality-floors
+autonomy: auto
+phase: done
+fast: true
+oneshot: true
+gate_mode: ai-plan-verify
+
+> Fast lane — one small task, minimal sections, filled top-to-bottom. The trust floor still
+> holds: a FROZEN §3 contract · ≥1 red test before build · a recorded §6 gate (security = HARD-STOP).
+> The acceptance scenario collapses into §1 `Accept:`; OBSERVE is one optional line at the gate.
+
+---
+
+## 0 · GROUND — the real codebase
+
+Touches (files · symbols): `add-method/tooling/templates/TASK.md.tmpl:127-129` (§6 checklist + `### Deep checks` block — the shallow audit already measures that block, so a dialect bullet there is audit-counted with ZERO new engine code) · `templates/TASK.fast.md.tmpl:79-80` (fast §6 checklist) · `templates/GLOSSARY.md.tmpl:5-15` (`## Sensitivity classes` — guidance must be PROSE, never a `- token:` bullet, or `_SENS_CLASS_LINE_RE` mints a new class) · `.add/GLOSSARY.md` (this project's own, section currently absent) · `sensitivity.md` skill guide ×3 twins · templates ×3 tooling twins; engine_manifest covers add_engine/*.py ONLY → no pin re-aim (add.py untouched)
+Context (working folder): quality-floors MILESTONE.md items 3+4; evidence = benchmark/results/2026-07-wv1-rep0.md wm2 root cause
+Honors (patterns / conventions): §6 already uses `<how / where>` placeholders (safe to mirror; the v16 tag census binds §1-§4 spans, not §6) · template/guide edits propagate to all twins before the gate · guidance-not-token (base `data` class, no new vocabulary)
+Anchors the contract cites: `TASK.md.tmpl ### Deep checks` · `TASK.fast.md.tmpl §6` · `GLOSSARY.md.tmpl ## Sensitivity classes` · `_project_sensitivity_domain` (must stay ()-stable) · `sensitivity.md`
+Ground SHA: `f9d2303`
+Skip rationale: scenarios — two template lines + two prose entries, §1 Accept covers; observe — one optional delta line at the gate
+
+---
+
+## 1 · SPECIFY — the rules
+
+Feature: quality-floors levers 3+4 — a §6 input-dialect check line in BOTH task templates (audit-counted via the existing Deep-checks/shallow machinery) + the datetime/money/timezone ⇒ `data` sensitivity guidance in the GLOSSARY template, this project's GLOSSARY, and the sensitivity skill guide (wm2 evidence named)
+Must:
+  - TASK.md.tmpl `### Deep checks` gains `- [ ] DIALECT — the tests speak the same value formats as the spec's own examples (spec-dialect floor): <what confirmed>` — inside the block the `shallow` audit already measures
+  - TASK.fast.md.tmpl §6 checklist gains the condensed twin line
+  - GLOSSARY.md.tmpl `## Sensitivity classes` gains PROSE guidance (not a bullet): datetime/money/timezone arithmetic ⇒ declare `data` (full lane), wm2 evidence in one clause; `.add/GLOSSARY.md` gains the same section content for THIS project
+  - sensitivity.md skill guide names the rule + the wm2 evidence; all template/guide twins byte-identical
+  - the guidance mints NO new sensitivity token — `_project_sensitivity_domain` over a fresh render stays ()
+Reject:
+  - guidance written as a `- token:` bullet -> "vocab_leak" (a new class token would silently widen freeze/status/check vocabulary)
+  - a §1-§4 template span gaining a bare angle-tag -> "tag_census_collision" (v16 hazard; §6 only)
+Accept: Given the templates render a fresh task, When §6 is read, Then both templates carry the dialect line (full inside Deep checks) AND the GLOSSARY template + project GLOSSARY + sensitivity guide carry the datetime/money/tz ⇒ data rule AND `_project_sensitivity_domain` of a fresh render is unchanged ()
+Assumptions: ⚠ template guard suites may pin §6 line counts or exact Deep-checks content — if wrong: those pins red on the new line and each gets a TESTS-re-cross amendment in ITS owning suite (cost: one re-cross loop)
+
+---
+
+## 3 · CONTRACT — freeze the shape
+
+```
+TASK.md.tmpl / ### Deep checks (audit-counted via `shallow`):
+  - [ ] DIALECT — the tests speak the same value formats as the spec's own examples
+        (spec-dialect floor): <what confirmed>
+TASK.fast.md.tmpl / §6 checklist:
+  - [ ] input dialect held — tests speak the spec's example formats (spec-dialect floor)
+GLOSSARY.md.tmpl + .add/GLOSSARY.md / ## Sensitivity classes (PROSE, no bullet):
+  "Datetime, money, or timezone arithmetic ⇒ declare `data` …(wm2 evidence clause)"
+sensitivity.md (×3 skill twins): the same rule + evidence, one short block.
+success: fresh render carries the lines; _project_sensitivity_domain stays ().
+rejections: vocab_leak (no `- token:` guidance) · tag_census_collision (§6 only).
+```
+
+`Least-sure flag surfaced at freeze:` [test] existing template-guard suites may pin §6/Deep-checks content or counts — the build may red THEIR pins, and each amendment must go through a TESTS re-cross in the owning suite, never a quiet edit; cost if many pin: several re-cross loops
+Status: FROZEN @ v1 — approved by claude-fable-5
+Freeze mode: ai-plan-verify — verified by claude-fable-5 at 2026-07-10T17:38:23+00:00
+
+### AI-verify record (required when gate_mode: ai-plan-verify)
+- [x] §0 GROUND anchors resolve in the current tree — template line numbers, GLOSSARY tmpl section, sensitivity.md ×3, _SENS_CLASS_LINE_RE reader all grepped at f9d2303
+- [x] §1 every Must + every Reject present, each Reject paired with an error code (vocab_leak · tag_census_collision)
+- [x] §3 CONTRACT shape is concrete — exact line text per surface
+- [x] Lowest-confidence flag surfaced and substantive — template-guard pin collisions with the re-cross cost
+Verified by: claude-fable-5 (session ee9aef91, orchestrator inline) · at: 2026-07-10T21:30:00Z
+
+---
+
+## 4 · TESTS — failing-first (red)
+
+Plan: test_dialect_vocab_lines.py — test_full_template_deep_checks_carries_dialect_line · test_fast_template_carries_dialect_line · test_glossary_template_carries_data_guidance_as_prose (+ regex proves no `- token:` bullet added) · test_project_glossary_carries_the_rule · test_sensitivity_guide_names_wm2_evidence · test_no_new_sensitivity_token (render a fresh project, _project_sensitivity_domain == ()).
+Tests live in: `add-method/tooling/test_dialect_vocab_lines.py` · MUST run red before Build.
+
+---
+
+## 5 · BUILD — AI writes code
+
+Scope (may touch): `add-method/` `.claude/` `.add/GLOSSARY.md` `tmp/`
+Strategy & known-problem fixes: 1. red pins 2. canonical template/guide/glossary edits 3. cp to every twin 4. full tooling suite; traps: guidance as prose only (vocab_leak) · §6 only (tag census) · twins before the gate
+Approach (domain strategy): book-technical-writer stance — the floor's prose surface: one checklist line a verifier can act on, one glossary rule a task author can route by
+Strategy actually used: as planned, plus the byte-ceiling absorption the lean rule demands — the DIALECT line and sensitivity.md guidance were compressed (all §3-cited phrases held, pinned by the red suite) and ~183B of existing sensitivity.md prose trimmed so pool 51885/tree 145974 fences hold; 5 sibling pins amended via TESTS re-cross exactly as the least-sure flag predicted (4 Deep-checks fixtures now fill the DIALECT placeholder; fast-lane's task-local empty-git-diff guard re-pinned to its durable invariant: skip machinery stays fast-lane-only); a 4th gitignored dogfood twin (add-method/.add/tooling/templates/) joined the sync set.
+Code lives in: `./src/`   ·   Constraints: change no test, no contract; allow-list packages only.
+
+---
+
+## 6 · VERIFY — evidence + gate
+
+- [x] all tests pass · coverage held · no test or contract altered during build — full tooling suite 3374 passed + 162 subtests, 0 failed; test amendments (5) went through an explicit TESTS re-cross, never a quiet edit
+- [x] green was EARNED — the 6 red pins assert rendered template/guide CONTENT (phrase + placement + no-bullet regex), not the build's own strings; `_project_sensitivity_domain` () proven over a fresh render
+- [x] no exposed secrets, injection openings, or unexpected dependencies — prose/template-only change; no new imports, no engine code touched (ENGINE_MD5/PKG unchanged)
+
+Build expectations (from §1 Accept + §3 CONTRACT): a fresh render carries the DIALECT line inside `### Deep checks` (full) and the condensed §6 line (fast); GLOSSARY.md.tmpl + `.add/GLOSSARY.md` + sensitivity.md ×3 carry datetime/money/tz ⇒ `data` with the wm2 evidence; no new sensitivity token — confirmed by test_dialect_vocab_lines.py 6/6 + budget/parity/audit suites green (3374 total).
+
+### GATE RECORD
+Outcome: PASS
+Reviewed by: Tin Dang · date: 2026-07-11
+[SPEC · open] lever-1 refinement: `_dialect_gaps` scans the §3 RAW body, which includes engine-written freeze metadata — this task's own tests→build crossing warned `aware-iso-timestamp` on its `Freeze mode: … 2026-07-10T17:38:23+00:00` stamp, a false positive of the v1 scan span; candidate fix: scan only §3 fenced blocks or strip the Status/Freeze-mode lines (evidence: crossing output at the phase-build re-cross, this task, 2026-07-11)
+