scbe-aligned-foundations

star 6

Use when building, extending, or auditing the SCBE multi-representation training program that aligns mathematics, English, Sacred Tongues lane naming, binary transport framing, chemistry packets, and coding primaries into one staged curriculum. Trigger this skill for tokenizer-native language lanes, chemistry-as-structure training, coding-primaries alignment, aligned foundation dataset generation, or transfer-eval planning across those lanes.

issdandavis By issdandavis schedule Updated 4/26/2026

name: scbe-aligned-foundations description: >- Use when building, extending, or auditing the SCBE multi-representation training program that aligns mathematics, English, Sacred Tongues lane naming, binary transport framing, chemistry packets, and coding primaries into one staged curriculum. Trigger this skill for tokenizer-native language lanes, chemistry-as-structure training, coding-primaries alignment, aligned foundation dataset generation, or transfer-eval planning across those lanes.


Use this skill when the task is to keep the SCBE token substrate aligned across multiple representations instead of treating each dataset as an isolated fine-tune.

Core rule

Preserve one concept across multiple synchronized forms:

  • mathematics
  • plain English
  • Sacred Tongues abbreviation and full-name lane
  • binary or packet framing
  • domain lane instantiation such as chemistry or coding

Do not collapse those into one flat corpus if the alignment can be preserved.

Workflow

  1. Build or refresh the aligned foundations set:
    • python scripts/build_aligned_foundations_sft.py
  2. Build or refresh the chemistry lane:
    • python scripts/build_chemistry_primary_sft.py
  3. Treat local coding corpora as first-class:
    • training-data/sft/bijective_codeflow_v1_*.sft.jsonl
    • training-data/sft/drill_langues_full_*.sft.jsonl
  4. Use the combined profile when the goal is shared substrate learning:
    • config/model_training/aligned-foundations-qwen-primary.json
  5. Use the chemistry-only profile when isolating stability or conservation:
    • config/model_training/chemistry-qwen-primary.json

What to check before training

  • Full tongue names and abbreviations co-occur often enough
  • Layer numbers and names co-occur for L10 through L14
  • Risk tiers are taught as a full set: ALLOW, QUARANTINE, ESCALATE, DENY
  • Chemistry records state why a reaction is stable or unstable
  • Coding records preserve slot alignment and bijective mappings

Evaluation priorities

Judge success by structure, not just loss:

  • token meaning stability
  • cross-lane semantic preservation
  • packet format compliance
  • chemistry conservation and stability correctness
  • coding slot and algorithm preservation
  • transfer between chemistry, coding, and governance language

References

Read these when shaping or auditing the lane:

Install via CLI
npx skills add https://github.com/issdandavis/SCBE-AETHERMOORE --skill scbe-aligned-foundations
Repository Details
star Stars 6
call_split Forks 1
navigation Branch main
article Path SKILL.md
More from Creator