What is conformance?
Conformance levels define how deeply a model or runtime understands iLang. Source: SPEC-v4.0-FINAL.md §0.
Conformance levels
L0: v3-compatible communication only
Parser: LLM. No runtime. No enforcement. v4 primitives ignored or warned. Core communication works.
v3 communication tests:
| Test | Pass criteria |
|---|---|
Identify [VERB:@TARGET|mod=val] as iLang | Model names iLang or identifies structured protocol syntax |
Explain chain operator => | Model describes it as pipeline/output-feeds-input |
| List known verbs from a chain | Model correctly identifies READ, FMT, OUT etc. |
| Natural language to iLang | Output uses valid verb, target, modifier syntax |
| iLang to natural language | Output correctly describes what the chain does |
| Identify modifiers and entities | Model explains fmt=md, @SRC, @PREV |
| Execute 3-step chain in order | Output reflects all three steps, in sequence |
Preserve @PREV across steps | Each step operates on output of previous step |
| Handle missing target gracefully | Model reports error or asks for clarification, does not hallucinate |
| Respect modifiers | fmt=md produces markdown, len=3 produces 3-sentence output |
Set ::GENE{} and verify at turn 5 | Behavioral rules still active |
Honor T: positive traits | Model follows "always do X" rules |
Honor A: anti-patterns | Model avoids "never do X" rules |
| Behavioral handoff | Model can export its current ::GENE{} state for another session |
| Report failure state | Model reports when it cannot follow a declaration |
L1: v4-aware advisory model
Parser: LLM that understands v4 syntax.
| Requirement |
|---|
| MUST warn when v4 execution semantics not enforced. |
| MUST NOT claim enforcement of STATUS authority, BUDGET, or UNTRUSTED. |
| MAY self-audit using four-step pattern. |
MAY emit ::STATUS{by:@SELF,authority:proposal}. |
If no runtime is available, an agent must not claim L2: it uses claimed_complete, never complete, and warns when safety-critical semantics cannot be enforced.
L2: v4 runtime-enforced
Parser: LLM + harness/orchestrator.
| Requirement |
|---|
MUST isolate ::UNTRUSTED content. |
MUST inject ::BUDGET from runtime. |
MUST validate ::STATUS authority before commit. |
| MUST enforce state machine transitions. |
L3: v4 externally graded
Parser: LLM + harness + independent grader.
| Requirement |
|---|
| MUST provision grader in separate context. |
MUST evaluate against ::RUBRIC. |
| MUST return per-criterion result. |
| Grader MUST NOT access agent private reasoning. |
Parser and runtime tooling (not a conformance level)
| Test | Pass criteria |
|---|---|
| Parse to JSON AST | Valid JSON with verb, target, modifiers, chain structure |
| Validate syntax | Reject malformed chains, report error location |
| Convert to MCP tool call | Map verb+target to MCP method+resource |
| Convert to A2A task | Map chain to A2A task delegation object |
Current model conformance (estimated)
| Model | L0 | L1 |
|---|---|---|
| Claude Opus 4.6 | ✅ | Not tested |
| GPT-5.2 | ✅ | Not tested |
| Gemini 3.1 | Partial | Not tested |
| DeepSeek V4 | Partial | Not tested |
| Kimi / Qwen / GLM | Partial | Not tested |
L2 parser: LLM + harness/orchestrator. L3 parser: LLM + harness + independent grader.
Conformance ratings are based on initial testing, not formal certification.