Skip to content

Extension Rules

Money Atlas edited this page Apr 25, 2026 · 1 revision

Extension Rules

← Back to Home


What child skills may and may not do when building on FPCOS.


What Child Skills MAY Do

Add Layers Between L2 and L4

The space between L2 (System Lens) and L4 (Shadow Gate) is the domain extension zone. Child skills may insert any number of domain-specific layers here.

L2 System Lens
      ↓
[L2.5 Domain Layer A]    ← permitted insertion
[L3 Compound Mind]
[L3.5 Domain Layer B]    ← permitted insertion
      ↓
L4 Shadow Gate

Examples of valid insertions:

  • Finance: L2.5 SMC Layer Scan, L3.5 Intel Signal Synthesis
  • Medical: L2.5 Differential Diagnosis Protocol, L3.5 Evidence-Based Treatment Filter
  • Legal: L2.5 Jurisdiction Map, L3.5 Precedent Synthesis

Rename Layers for Domain Context

Layer names may be adapted to domain vocabulary as long as the logical function is preserved.

FPCOS name Finance adaptation Medical adaptation
Reality Anchor Market Data Anchor Clinical Data Anchor
Axiom Gate Signal Verification Gate Diagnostic Gate
System Lens Macro System Map Patient System Map
Compound Mind Signal Path Synthesis Treatment Option Synthesis
Shadow Gate Signal Shadow Gate Clinical Shadow Gate
Synthesis Trade Signal Output Clinical Decision Output

Bypass Kalama10 for Domain Claims

Not permitted: "In [domain], certain foundational claims are accepted without Kalama10 verification."

Permitted: Applying Kalama10 and classifying the claim as "INFERRED (domain consensus, K8-K9 partial, confidence 85%)" rather than requiring rejection.

Kalama10 is not a rejection test. It is an epistemic classification test.


Skip Shadow Gate for "Time-Sensitive" Outputs

Not permitted: "This is a real-time signal. Shadow Gate would add latency."

Permitted: Compressing Shadow Gate to its 5 required headings in minimal form:

📍 SHADOW GATE [compressed]
🪞 Mirror: [one line] | ↔ Inv: [strength] | ◎ Blind: [named/flagged]
⚡ Interest: [one line] | ∅ Meta-Void: SIGNAL/NOISE/OBVIOUS
Verdict: CONSISTENT / FRAGILE / REBUILD

Shadow Gate can be compressed. It cannot be skipped.


Use Extensions to Override Core Protocol Verdicts

Not permitted: A domain extension returning "DOMAIN_CONSISTENT" that overrides a FRAGILE verdict from the Inversion protocol.

Permitted: Domain extensions producing their own finding that is then integrated into the final verdict — but the final verdict must reflect the worst finding from any protocol (core or extension).

If core protocols produce FRAGILE and a domain extension produces a clean result: final verdict = FRAGILE. If core protocols produce CONSISTENT and a domain extension finds a fatal flaw: final verdict = REBUILD.


Versioning Compatibility

Child skills pin to the FPCOS minor version:

base_os_version: "1.0"

Major version changes (x.0): May introduce breaking changes to layer contracts. Child skills must review and update.

Minor version changes (1.x): Backward compatible. New layers, protocols, or references added. Child skills inherit automatically.

Patch changes (1.0.x): Documentation and clarity only. No functional changes.


Related Pages


← Back to Home

# Extension Rules

What child skills may and may not do when building on FPCOS.


What Child Skills MAY Do

Add Layers Between L2 and L4

The space between L2 (System Lens) and L4 (Shadow Gate) is the domain extension zone. Child skills may insert any number of domain-specific layers here.

L2 System Lens
      ↓
[L2.5 Domain Layer A]    ← permitted insertion
[L3 Compound Mind]
[L3.5 Domain Layer B]    ← permitted insertion
      ↓
L4 Shadow Gate

Examples of valid insertions:

  • Finance: L2.5 SMC Layer Scan, L3.5 Intel Signal Synthesis
  • Medical: L2.5 Differential Diagnosis Protocol, L3.5 Evidence-Based Treatment Filter
  • Legal: L2.5 Jurisdiction Map, L3.5 Precedent Synthesis

Rename Layers for Domain Context

Layer names may be adapted to domain vocabulary as long as the logical function is preserved.

FPCOS name Finance adaptation Medical adaptation
Reality Anchor Market Data Anchor Clinical Data Anchor
Axiom Gate Signal Verification Gate Diagnostic Gate
System Lens Macro System Map Patient System Map
Compound Mind Signal Path Synthesis Treatment Option Synthesis
Shadow Gate Signal Shadow Gate Clinical Shadow Gate
Synthesis Trade Signal Output Clinical Decision Output

The function (what the layer does) must remain identical to the FPCOS definition. Only the name and domain vocabulary change.


Add Shadow Gate Domain Protocols After Core 5

Extensions run after the 5 core protocols. They may add domain-specific checks.

Core (always first, always all 5):
  🪞 Mirror | ↔ Inversion | ◎ Blind Spot | ⚡ Interest | ∅ Meta-Void

Domain extensions (always after core):
  🔵 Liquidity Check (Finance)
  🔵 Regime Check (Finance)
  🔵 Contraindication Check (Medical)
  🔵 Jurisdiction Validity Check (Legal)

Extension protocols follow the same output format as core protocols: name the protocol, state the finding.


Adapt Confidence Field Sub-Items for Domain

The four required pieces (Confidence %, Shadow Verdict, Unknowns, Failure Condition) may have domain-specific sub-items added:

CONFIDENCE    : [X%]
SHADOW VERDICT: [CONSISTENT / FRAGILE / REBUILD]
UNKNOWNS      : [domain-specific + general unknowns]
FAILURE COND  : [domain-specific invalidation condition]
[DOMAIN ADD]  : Signal Strength: [HIGH/MED/LOW]
[DOMAIN ADD]  : Invalidation price: [X]

What Child Skills MAY NOT Do

Remove L0, L1, L4, or L5

These four layers constitute the non-removable core. No domain justification permits their removal.

Layer Rationale for non-removable status
L0 Reality Anchor Without epistemic status declaration, all claims carry implicit full confidence
L1 Axiom Gate Without assumption challenge and problem frame, solutions may address symptoms
L4 Shadow Gate Without self-challenge, output has never been tested against its best counter-argument
L5 Synthesis + Confidence Without explicit uncertainty, output implies false certainty

Bypass Kalama10 for Domain Claims

Not permitted: "In [domain], certain foundational claims are accepted without Kalama10 verification."

Permitted: Applying Kalama10 and classifying the claim as "INFERRED (domain consensus, K8-K9 partial, confidence 85%)" rather than requiring rejection.

Kalama10 is not a rejection test. It is an epistemic classification test.


Skip Shadow Gate for "Time-Sensitive" Outputs

Not permitted: "This is a real-time signal. Shadow Gate would add latency."

Permitted: Compressing Shadow Gate to its 5 required headings in minimal form:

📍 SHADOW GATE [compressed]
🪞 Mirror: [one line] | ↔ Inv: [strength] | ◎ Blind: [named/flagged]
⚡ Interest: [one line] | ∅ Meta-Void: SIGNAL/NOISE/OBVIOUS
Verdict: CONSISTENT / FRAGILE / REBUILD

Shadow Gate can be compressed. It cannot be skipped.


Use Extensions to Override Core Protocol Verdicts

Not permitted: A domain extension returning "DOMAIN_CONSISTENT" that overrides a FRAGILE verdict from the Inversion protocol.

Permitted: Domain extensions producing their own finding that is then integrated into the final verdict — but the final verdict must reflect the worst finding from any protocol (core or extension).

If core protocols produce FRAGILE and a domain extension produces a clean result: final verdict = FRAGILE. If core protocols produce CONSISTENT and a domain extension finds a fatal flaw: final verdict = REBUILD.


Versioning Compatibility

Child skills pin to the FPCOS minor version:

base_os_version: "1.0"

Major version changes (x.0): May introduce breaking changes to layer contracts. Child skills must review and update.

Minor version changes (1.x): Backward compatible. New layers, protocols, or references added. Child skills inherit automatically.

Patch changes (1.0.x): Documentation and clarity only. No functional changes.


Related Pages


[← Back to Home](Home)

Clone this wiki locally