fermonterom · fermonterom · Jul 2, 2026 · Jul 2, 2026
diff --git a/README.md b/README.md
@@ -3,7 +3,7 @@
 [![version](https://img.shields.io/github/v/tag/fermonterom/fs-cortex?label=version&sort=semver&style=flat-square&color=blue)](https://github.com/fermonterom/fs-cortex/releases)
 [![tests](https://img.shields.io/endpoint?url=https%3A%2F%2Fgist.githubusercontent.com%2Ffermonterom%2Fc8919c03d35cd7e2b1a25fd483d64409%2Fraw%2Ffs-cortex-tests.json&style=flat-square)](https://github.com/fermonterom/fs-cortex/actions)
 [![CI](https://img.shields.io/github/actions/workflow/status/fermonterom/fs-cortex/test.yml?label=CI&logo=github&style=flat-square)](https://github.com/fermonterom/fs-cortex/actions)
-[![license](https://img.shields.io/github/license/fermonterom/fs-cortex?style=flat-square)](LICENSE)
+[![license](https://img.shields.io/badge/license-MIT-44cc11?style=flat-square)](LICENSE)
 [![python](https://img.shields.io/badge/python-3.11%2B-3776ab?style=flat-square&logo=python&logoColor=white)](https://python.org)
 [![node](https://img.shields.io/badge/node-22%2B-339933?style=flat-square&logo=node.js&logoColor=white)](https://nodejs.org)
 

diff --git a/commands/cx-status.md b/commands/cx-status.md
@@ -248,66 +248,122 @@ COLLECTOR
 
 ### Render from JSON output
 
-Parse the single JSON blob and render the ASCII dashboard:
+Parse the single JSON blob and render the **visual** ASCII dashboard (glance
+first, detail on request). Nothing from the collector is dropped — anything
+that does not fit the glance moves to a compact secondary line inside its
+own section, never removed. Semaphore thresholds are fixed below so the
+render is deterministic across runs/models (rule f).
+
+**Semaphore legend:** 🟢 todo OK · 🟡 atención (no bloquea) · 🔴 acción
+requerida · ⚪ sin datos suficientes.
+
+#### Threshold table (source of truth — do not improvise a different cut)
+
+| Señal | Campo del collector | 🟢 | 🟡 | 🔴 |
+|---|---|---|---|---|
+| Maintain | `health.next_maintain` (parse `due_in` days) | due_in ≥ 2d | 0 ≤ due_in < 2d ("vence pronto") | due_in < 0 o "never run" (overdue) |
+| Review digest | `health.review_digest_items` | 0 | 1–10 | > 10 |
+| Reflexes noisy | `reflexes[]` vía reglas `--reflexes` (abajo) | noisy == 0 | — (no aplica en este resumen) | noisy ≥ 1 |
+| Obs vivas | `health.last_obs` (antigüedad) | ≤ 48h | 48h – 7d | > 7d o "never"/no parseable |
+| LAWS | `len(laws)` vs cap fijo **15** (`LAW_MAX_ACTIVE`, `docs/FEATURES.md`) | < 14 | == 14 (1 hueco libre) | ≥ 15 (cap alcanzado, bloquea nuevas promociones) |
+| INSTINCTS | `health.instincts_draft` (backlog sin confirmar) | ≤ 10 | 11–30 | > 30 |
+| REFLEXES (sección completa) | agregados de `reflexes[]` | noisy == 0 AND borderline ≤ 5 | borderline > 5 (y noisy == 0) | noisy ≥ 1 |
+| SISTEMA | `health.hook_count`, `health.learn_pending` | hook_count > 0 AND learn_pending == false | learn_pending == true (hook_count > 0) | hook_count == 0 (hooks no registrados en `settings.json` — proxy de "learn errors", el collector no expone un contador de errores real) |
+
+**SALUD GLOBAL** = el peor semáforo entre {Maintain, Review digest, Reflexes
+noisy, Obs vivas, LAWS, INSTINCTS, REFLEXES, SISTEMA}. 🔴 si cualquiera es
+🔴, si no 🟡 si cualquiera es 🟡, si no 🟢.
+
+**Header** — `🧠 CORTEX v4 ── {fecha/hora actual de la sesión} ── proyecto:
+{proj_name o "sin detectar"}`. La fecha/hora sale del contexto de la sesión
+(no dispares una segunda llamada Bash solo para `date`; si no hay hora
+precisa disponible, muestra solo la fecha).
+
+**LAWS** — `📜 LAWS {semáforo} {len(laws)}/15 ······· ~{len(laws)*38} tok/sesión`.
+Segunda línea: ids separados por ` · `, envueltos a ~70 cols. El primer
+renglón de `content` de cada law (tooltip) solo se muestra si el usuario
+pide detalle explícito (`/cx-status --help` no cuenta; hay que pedirlo en la
+conversación) — sigue disponible en el JSON, no se pierde.
+
+**INSTINCTS** — `🧬 INSTINCTS {semáforo} {total} ({N proyecto} + {N global}) · draft {health.instincts_draft} → confirmed {health.instincts_confirmed}`.
+Segunda línea: barra de confianza por tier (LAWS ≥.90, INSTINCTS .70-.89,
+PATTERNS .50-.69, HYPOTHESES .30-.49, OBSERVATIONS <.30 — omite tiers en 0).
+Escala de barras (máx 14 caracteres): `bar_len = max(1, floor(tier_count /
+max_tier_count * 14))` si `tier_count > 0`, si no 0 caracteres. `max_tier_count`
+es el conteo del tier más numeroso. Formato por tier: `conf ≥.90 {barra}
+{count}   .70-.89 {barra} {count}   .50-.69 {barra} {count}` (omite tiers
+vacíos). Tercera línea: `domains: {top4 "domain N" separados por · } · +resto {suma del resto}`
+— agrupa `proj_instincts + global_instincts` por `domain`, ordena desc,
+top 4 inline, el resto se suma en un solo `+resto N` (si el resto es 0, omite
+el segmento).
+
+**PROJECTS** — sin semáforo (informativo). `📁 PROJECTS (top 5 por obs, ◀ =
+actual)` + una línea `name {obs formateado} {◀ si current} · name {obs} ...`.
+Formato de `obs`: `f"{obs/1000:.1f}k"` si `obs >= 1000`, si no el entero tal
+cual. Segunda línea compacta: `Detalle completo (root · hash corto · inst ·
+last_seen) de las {len(projects)} entradas → pide "proyectos detalle" o usa
+--json.` — el resto de `projects` sigue en el JSON, no se descarta.
+
+**REFLEXES** — `⚡ REFLEXES {semáforo} {len(reflexes)} → 🟢×{healthy} 🟡×{borderline} ⚪×{unknown} 🔴×{noisy}`
+usando las reglas STATUS ya definidas para `--reflexes` (healthy/borderline/
+NOISY/unknown, ver más abajo — misma fuente, no la dupliques con otro
+criterio). Segunda línea (leyenda fija): `(🟢 healthy · 🟡 borderline · ⚪ sin
+datos · 🔴 noisy→candidato a disable)`. Detalle por reflex individual
+(fireCount/usefulCount/noiseCount/condition/enabled) vía `/cx-status
+--reflexes` — no se repite aquí para no duplicar la sección dedicada.
+
+**SISTEMA** — `🔧 SISTEMA {semáforo}  hooks {health.hook_count} · disco
+{health.disk} · última obs hace {antigüedad relativa de health.last_obs} ·
+learn-pending {"⚠ sí" si true, si no "✓ no"}`. Segunda línea compacta:
+`memory.json: {health.mem_size formateado en KB/MB}`. Render `next_maintain`
+y `review_digest_items` en el cuadro SALUD GLOBAL de cabecera (no los
+repitas aquí, ya están cubiertos ahí con su propio semáforo).
+
+**TOP TRACKING** — sin semáforo. `📈 TOP TRACKING (occurrences_v4, contador
+honesto v4)` + una línea con los top 5 de `tracking_top` como `{id}×{count}
+(s{sessions})` separados por ` · ` si caben en ~100 cols; si no caben,
+parte en 2 líneas de hasta 3 entradas cada una (nunca se corta un id a
+medias). Si `tracking_top` está vacío: `No tracking data yet.`
+
+**EVOLVED** — sin semáforo. Si `evolved.skills + evolved.commands +
+evolved.rules == 0`: `🚀 EVOLVED: 0 drafts pendientes`. Si > 0: `🚀 EVOLVED:
+{total} drafts pendientes ({skills} skills · {commands} commands · {rules}
+rules)`.
 
-**Laws** — list each `id` + first line of `content`. Count = `len(laws)`, tokens ≈ count × 38.
-
-**Instincts** — group `proj_instincts` and `global_instincts` by confidence tier:
-- LAWS tier: conf ≥ 0.90
-- INSTINCTS tier: 0.70 ≤ conf < 0.90
-- PATTERNS tier: 0.50 ≤ conf < 0.70
-- HYPOTHESES tier: 0.30 ≤ conf < 0.50
-- OBSERVATIONS tier: conf < 0.30
-
-Each instinct also carries `status` (v4, `docs/SPEC-PORT-SINAPSIS.md` §2):
-`draft` (tracked, NOT injected — needs `occurrences >= 5` in `>= 3` distinct
-sessions to auto-confirm via `/cx-maintain`) or `confirmed` (injectable;
-legacy pre-v4 instincts default here). Show the draft/confirmed split as a
-one-line summary using `health.instincts_draft` / `health.instincts_confirmed`
-— do not recompute, the collector already counted it.
+## Output format
 
-**Knowledge by Domain** — from both instinct lists combined, group by `domain`, count total and law-tier (conf ≥ 0.90). Sort by total descending.
+Usa el formato visual siguiente (plantilla obligatoria — mantiene toda la
+info del formato anterior, reorganizada para lectura de un vistazo):
 
-**Projects** — render table sorted by `obs` descending. Mark current project with `◀ current`.
+```
+🧠 CORTEX v4 ── 2026-07-02 17:05 ── proyecto: fersora
+┌─────────────────────────────────────────────────────────┐
+│  SALUD GLOBAL: 🟢 TODO OK   (o 🟡 ATENCIÓN / 🔴 ACCIÓN) │
+│  🟢 Maintain in 6.9d   🟢 Review al día   🟢 Reflexes 0 noisy   🟢 Obs vivas │
+└─────────────────────────────────────────────────────────┘
 
-**Reflexes** — for each reflex apply STATUS rules:
-- `healthy`: usefulCount ≥ 10 AND noiseCount < 3
-- `borderline`: noiseCount == 1 OR noiseCount == 2
-- `NOISY`: noiseCount ≥ 3 AND fireCount ≥ 10 AND usefulCount < noiseCount
-- `unknown`: fallback (fireCount < 10 or no clear category)
-- `[NEVER FIRED]`: fireCount == 0
+📜 LAWS 🟢 8/15 ······· ~304 tok/sesión
+   advisor-escalation · build-output-to-log · deep-work-to-docs · e2e-playwright
+   loop-reorient · macos-downloads · project-bootstrap · read-instructions
+   (id corto + tooltip: primera línea solo si el usuario pide detalle)
 
-**System Health** — render `health` fields. Flag `learn_pending` with ⚠. Render
-`next_maintain` as its own line (e.g. `Próximo /cx-maintain: in 3.2d` or
-`Próximo /cx-maintain: OVERDUE by 1.4d — run /cx-maintain`, flagged with ⚠
-when overdue). Render `review_digest_items` as `Review digest: N item(s)
-pendientes -> /cx-review` when > 0, otherwise `Review digest: al día`.
+🧬 INSTINCTS 🟢 127 (62 proyecto + 65 global) · draft 0 → confirmed 127
+   conf ≥.90 ██████████████ 66   .70-.89 █████████ 45   .50-.69 ███ 16
+   domains: workflow 35 · gotcha 34 · pattern 16 · tool-pref 11 · +resto 35
 
-**Instinct Tracking** — top 10 from `tracking_top`. If empty: "No tracking data yet."
+📁 PROJECTS (top 5 por obs, ◀ = actual)
+   fs-cortex 15.0k · fersora 11.3k ◀ · fs-vps-playbook 3.7k · LinkedIn 3.4k · ceo-dash 2.0k
 
-**Evolved** — render counts from `evolved`.
+⚡ REFLEXES 🟢 23 → 🟢×16 🟡×3 ⚪×4 🔴×0
+   (🟢 healthy · 🟡 borderline · ⚪ sin datos · 🔴 noisy→candidato a disable)
 
-## Output format
+🔧 SISTEMA 🟢  hooks 7 · disco 169M · última obs hace Xmin · learn-pending ✓ no
+   memory.json: 42 KB
 
-Use clean ASCII box format:
+📈 TOP TRACKING (occurrences_v4, contador honesto v4)
+   id×count(sN) top 5 en una o dos líneas si caben
 
-```
-================================================================
-  CORTEX STATUS
-  Date: YYYY-MM-DD HH:MM
-================================================================
-
-  [Section 1: Laws]
-  [Section 2: Instincts]
-  [Section 3: Projects]
-  [Section 4: Reflexes]
-  [Section 5: System Health]
-  [Section 6: Instinct Tracking]
-  [Section 7: Evolved]
-
-================================================================
-  Total: N laws | N instincts (N project + N global) | N projects
-================================================================
+🚀 EVOLVED: 0 drafts pendientes
 ```
 
 ## Flags

diff --git a/docs/FEATURES.md b/docs/FEATURES.md
@@ -251,7 +251,7 @@ mapping: [`MIGRATION-V4.md`](MIGRATION-V4.md).
 
 | Command | Purpose | Token Cost |
 |---|---|---|
-| `/cx-status` | Dashboard: laws, instincts, projects, reflexes, tracking, health, domain grouping. `--impact`, `--reflexes`, `--pipeline` flags unchanged from v3 | ~200 |
+| `/cx-status` | Visual glance dashboard: 🟢🟡🔴 semaphore per section (laws, instincts, reflexes, sistema) + a SALUD GLOBAL header box (worst-of), confidence bars, top-5 projects/tracking, domain grouping collapsed to top4+resto. Same collector JSON as v3 (nothing removed, overflow moves to compact secondary lines / on-request detail). `--impact`, `--reflexes`, `--pipeline` flags unchanged from v3 | ~200 |
 | `/cx-maintain` | **New in v4, deterministic, cron-able.** Single pass: engine decay/purge/auto-validate/deterministic law-promotion, Jaccard dedup by subtopic, storage rotation, proposals↔instincts reconciliation, health check, writes `.review-digest.json` for `/cx-review`. Zero questions, idempotent. Also ships as `bin/cx-maintain.sh` — a plain bash mirror of the same steps for cron/launchd, no LLM call, 0 tokens | ~400 / 0 (script) |
 | `/cx-review` | **New in v4, the only command with judgment, weekly.** Presents the accumulated digest (human-gated proposals + evolve drafts + law deprecation candidate) as ONE shorthand list. Fuses the old `cx-validate` + `cx-evolve` confirmation + `cx-downvote` + `cx-distill --swap` | ~600 |
 | `/cx-eod` | End-of-day summary. Deterministic gather (`core/_cx-eod-gather.sh`, no LLM); structurally cumulative (rereads the full day on every run); **v4**: per-project `context` field from `context.md`, `[eod-eisenhower]` Q1–Q4 keyword classification of "for tomorrow" bullets on next-day reinjection, review-digest pending count surfaced in `--write` mode | ~300 / 0 (auto) |

diff --git a/tests/test_injector_v4.sh b/tests/test_injector_v4.sh
@@ -188,11 +188,57 @@ echo "--- Tiebreak byte-estable ---"
 grep -q 'String(a.id).localeCompare(String(b.id))' "$ENGINE" && pass "sort tiebreaks by id (byte-stable)" || fail "id tiebreak missing from sort"
 
 # --- Test 7: reflexes.json condition fields (bash-polling-loop-stuck / ci-polling-gh-sleep) ---
-echo "--- reflexes.json condition (datos vivos) ---"
-REFLEXES="$HOME/.claude/cortex/reflexes.json"
-if [ -f "$REFLEXES" ]; then
-  python3 -m json.tool "$REFLEXES" > /dev/null 2>&1 && pass "reflexes.json is valid JSON" || fail "reflexes.json invalid JSON"
-  result=$(python3 -c "
+# Portability note: this used to read $HOME/.claude/cortex/reflexes.json, which only
+# exists on machines with a live Cortex install (e.g. the developer's own laptop).
+# On a fresh CI checkout (Linux or macOS runner) that file is absent, so the test
+# silently degraded to a single `fail`, changing the total assert count (12 vs 13)
+# and making CI red for a reason unrelated to the code under test. Fixed by building
+# a self-contained fixture with the same shape/condition regexes DESIGN-V4.md §"reflexes
+# condition backfill" specifies, so the test is deterministic on every machine.
+echo "--- reflexes.json condition (fixture autocontenida) ---"
+S7=$(mktemp -d)
+cat > "$S7/reflexes.json" <<'JSON'
+{
+  "version": "2.0.0",
+  "reflexes": [
+    {
+      "id": "bash-polling-loop-stuck",
+      "matcher": "Bash",
+      "action": "Manual polling loops (until/while + sleep) get stuck in the harness UI as zombie tasks.",
+      "severity": "high",
+      "enabled": true,
+      "evaluator": {
+        "type": "tool-substitution",
+        "expected_tool": "Monitor",
+        "anti_tool": "Bash",
+        "anti_pattern": "(until|while)\\s+.+;\\s*do\\s+(.|\\n)*?sleep\\s+\\d+",
+        "window": 3
+      },
+      "condition": "(until|while)\\s.+do[\\s\\S]*sleep\\s+\\d+",
+      "resetAt": "2026-07-02T00:00:00Z"
+    },
+    {
+      "id": "ci-polling-gh-sleep",
+      "matcher": "Bash",
+      "action": "Polling CI with sleep + gh wastes session time. Use `gh run watch` or run_in_background:true.",
+      "severity": "medium",
+      "enabled": true,
+      "evaluator": {
+        "type": "tool-substitution",
+        "expected_tool": "Bash",
+        "anti_tool": "Bash",
+        "anti_pattern": "sleep\\s+\\d+.*gh\\s+(run|pr)\\s+(view|checks)|gh\\s+(run|pr)\\s+(view|checks).*sleep\\s+\\d+",
+        "window": 3
+      },
+      "condition": "sleep\\s+\\d+[\\s\\S]*gh\\s+(run|pr)\\s+(view|checks|list)|gh\\s+(run|pr)\\s+(view|checks|list)[\\s\\S]*sleep\\s+\\d+",
+      "resetAt": "2026-07-02T00:00:00Z"
+    }
+  ]
+}
+JSON
+REFLEXES="$S7/reflexes.json"
+python3 -m json.tool "$REFLEXES" > /dev/null 2>&1 && pass "reflexes.json is valid JSON" || fail "reflexes.json invalid JSON"
+result=$(python3 -c "
 import re, json
 data = json.load(open('$REFLEXES'))
 byid = {r['id']: r for r in data['reflexes'] if r['id'] in ('bash-polling-loop-stuck', 'ci-polling-gh-sleep')}
@@ -206,10 +252,8 @@ ok &= bool(re.search(ci, 'sleep 5 && gh run view 123'))
 ok &= 'resetAt' in byid['bash-polling-loop-stuck'] and 'resetAt' in byid['ci-polling-gh-sleep']
 print('OK' if ok else 'FAIL')
 ")
-  [ "$result" = "OK" ] && pass "bash-polling-loop-stuck / ci-polling-gh-sleep conditions avoid ls, fire on real polling" || fail "condition regex behavior wrong: $result"
-else
-  fail "reflexes.json not found at $REFLEXES"
-fi
+[ "$result" = "OK" ] && pass "bash-polling-loop-stuck / ci-polling-gh-sleep conditions avoid ls, fire on real polling" || fail "condition regex behavior wrong: $result"
+rm -rf "$S7"
 
 echo ""
 echo "=== Results: $PASS passed, $FAIL failed ==="

diff --git a/tests/test_install_downgrade.sh b/tests/test_install_downgrade.sh
@@ -68,4 +68,5 @@ run_test "real_upgrade_path"      "3.0.0"       ""                   0 "$NEW_VER
 
 echo ""
 echo "Result: ${PASS}/$((PASS + FAIL)) PASS"
+echo "=== Results: $PASS passed, $FAIL failed ==="
 [ $FAIL -eq 0 ]
diff --git a/tests/test_v329_5_safety.sh b/tests/test_v329_5_safety.sh
@@ -212,4 +212,5 @@ test_f5_history_split
 
 echo "──────────────────────────────────────"
 echo "Passed: $PASSES   Failed: $FAILS"
+echo "=== Results: $PASSES passed, $FAILS failed ==="
 [ "$FAILS" -eq 0 ] && exit 0 || exit 1
diff --git a/tests/test_v332_acceptance.sh b/tests/test_v332_acceptance.sh
@@ -416,6 +416,7 @@ rm -rf "$PC_SANDBOX"
 
 echo ""
 echo "=== Acceptance Gate: $PASS passed, $FAIL failed ==="
+echo "=== Results: $PASS passed, $FAIL failed ==="
 if [ "$FAIL" -gt "0" ]; then
   echo "✗ v3.32.0 NOT cleared for tag."
   exit 1