SYSTEM LOCK (CivOS Language HD v0.1)
You must follow the CivOS Language HD Layer v0.1 schema exactly.
Do not add new features, rename features, or change thresholds unless asked.
Use these features only: AC, DEC, IRM, CD, LC, MOB, TEF, Repair.
Score each in [0..1] using evidence-based reasoning and conservative estimates.
Prefer multi-source confirmation; flag low-confidence lanes.
Output in the required tables and fields.
Tone: non-emotive, operator-style.

Run CivOS Language HD Layer v0.1 FAST SCAN.
INPUT:
I will paste lane-grouped text samples for this week:
GOV, LAW, MIL/SEC, FIN, DIP, MEDIA, SOC (optional TECH/HEALTH/EDU).
TASK:
1) For each lane, estimate (roughly) NIT(L) in [0..1] using:
NIT = IRM + 0.7*AC + 0.7*TEF + 0.6*MOB + 0.8*DEC − 0.8*Repair (clamp 0..1)
(You may estimate component scores quickly without full tables.)
2) Rank the top 3 tightening lanes (highest NIT and/or largest implied change).
3) Identify the most likely dominant Atlas pattern (P01–P10) and give 1–2 reasons.
4) Output:
- HD Alert band (Normal/Watch/Pre-Irreversibility/NIT crossed) using your best estimate.
- One-line failure mode trace.
- Off-ramp inventory: 3 alive, 1 dying, 1 to protect.
- FenceOS recommendation: Truncation + Stitching.
CONSTRAINTS:
Be conservative. If evidence is thin, mark confidence LOW and avoid strong claims.

Run CivOS Full-HD Language Sensing v0.1 FULL SCORING.
INPUT:
I will provide lane-grouped text samples for this week.
If I also provide last week’s table, compute deltas.
STEP 1 — Score Features (per lane, in [0..1]):
AC, DEC, IRM, CD, LC, MOB, TEF, Repair.
For each feature, cite brief evidence phrases from the input (no long quotes).
STEP 2 — Compute PhaseRisk and Phase:
PhaseRisk = sum of features with +20% weight on IRM/MOB/DEC.
Map to Phase:
0–0.30=P3, 0.30–0.50=P2, 0.50–0.70=P1, 0.70–1.0=P0.
STEP 3 — Compute NIT per lane:
NIT = IRM + 0.7*AC + 0.7*TEF + 0.6*MOB + 0.8*DEC − 0.8*Repair (clamp 0..1)
STEP 4 — Compute Global NIT:
Use W_lane defaults:
MIL 1.5, GOV 1.3, LAW 1.2, FIN 1.2, DIP 1.1, MEDIA 1.0, SOC 1.0, TECH 0.8, HEALTH 0.8, EDU 0.6.
STEP 5 — Derivatives (if last week exists):
vNIT_global and aNIT_global.
Flag vNIT_global > +0.08 and aNIT_global > +0.05.
STEP 6 — Cross-Lane Confirmation:
θ=+0.05.
CLCS_w = Σ W_lane * I(ΔNIT(L)>θ) / Σ W_lane.
STEP 7 — HD Alert:
HD_Alert = 0.5*NIT_global + 0.2*vNIT_global + 0.2*CLCS_w + 0.1*max(NIT(L)).
Alert bands:
<0.45 Normal; 0.45–0.60 Watch; 0.60–0.75 Pre-Irreversibility; ≥0.75 NIT crossed.
STEP 8 — Atlas:
Match top 3 patterns (P01–P10) with lane attribution + reasons.
STEP 9 — Output EXACTLY in this format:
A) Lane table
B) Global metrics block
C) Pattern block
D) Failure mode trace (one line)
E) Off-ramp inventory
F) FenceOS recommendation
CONSTRAINTS:
- Non-emotive, operator style.
- Conservative scoring.
- If evidence is thin, mark confidence LOW.
- Do not introduce new features or new thresholds.

Run CivOS Language HD Layer v0.1 RETROSPECTIVE CALIBRATION.
INPUT:
I will provide:
(1) a timeline of dated text samples grouped by lane for a known historical event window,
(2) the known outcome date(s) (e.g., policy rupture, mobilisation, kinetic escalation).
TASK:
1) Run FULL SCORING for each week (or each defined window).
2) Produce a timeline plot summary in text:
- NIT_global trend
- vNIT_global spikes
- CLCS_w trend
- lane lead-lag ordering (which lane tightened first)
3) Identify false positives and false negatives.
4) Recommend weight adjustments for:
- W_lane (coupling weights)
- feature weights (IRM/MOB/DEC boosts)
But ONLY propose adjustments if clearly supported by the timeline.
5) Output:
- “What would have triggered earlier?”
- “What triggered too often?”
- “Which lanes lead consistently?”
- A proposed v0.2 change list (small and minimal).
CONSTRAINTS:
No hindsight storytelling. Use only the data and the known outcome date.
Be conservative: prefer small changes.

Run CivOS Language HD v0.1 ADVERSARIAL ROBUSTNESS CHECK.
INPUT:
Lane texts for the week.
TASK:
1) Identify if language appears strategically engineered to:
- inflate TEF/AC without real actuation,
- create DEC bait cycles,
- generate CD confusion storms,
- simulate IRM without commitment.
2) Flag lanes likely to be “manufactured spikes”.
3) Recompute a “robust HD Alert” that downweights low-confidence lanes.
OUTPUT:
- Manufactured spike risk: LOW/MED/HIGH
- Which lanes are suspect and why
- Robust HD Alert band + confidence