Backbone
暂无简介
- License 许可
- MIT
- 所属仓库
- rhCat/skillChip
- 最近更新
- 2026-07-31
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- Backbone
- perks
- [ "validate", "tlc", "tlaps", "preflight" ]
- skill
- backbone
02 资源文件 32
- 参考文档blueprint.json3.8 KB
- 参考文档index.json3.4 KB
- 参考文档ledger.json276 B
- 参考文档perks.json804 B
- 参考文档perks/preflight/manifesto.json528 B
- 参考文档perks/preflight/metadata.json951 B
- 脚本perks/preflight/src/bb_preflight.py4.8 KB
- 脚本perks/preflight/src/bb_preflight.sh240 B
- 参考文档perks/preflight/src/contracts.json510 B
- 参考文档perks/preflight/test/case.json395 B
- 参考文档perks/preflight/test/fixture/good.blueprint.json846 B
- 参考文档perks/tlaps/manifesto.json439 B
- 参考文档perks/tlaps/metadata.json947 B
- 脚本perks/tlaps/src/bb_tlaps.py3.8 KB
- 脚本perks/tlaps/src/bb_tlaps.sh228 B
- 参考文档perks/tlaps/src/contracts.json424 B
- 参考文档perks/tlaps/test/case.json390 B
- 参考文档perks/tlaps/test/fixture/good.blueprint.json846 B
- 参考文档perks/tlc/manifesto.json435 B
- 参考文档perks/tlc/metadata.json911 B
- 脚本perks/tlc/src/bb_tlc.py1.7 KB
- 脚本perks/tlc/src/bb_tlc.sh222 B
- 参考文档perks/tlc/src/contracts.json414 B
- 参考文档perks/tlc/test/case.json367 B
- 参考文档perks/tlc/test/fixture/good.blueprint.json846 B
- 参考文档perks/validate/manifesto.json349 B
- 参考文档perks/validate/metadata.json805 B
- 脚本perks/validate/src/bb_validate.py1.4 KB
- 脚本perks/validate/src/bb_validate.sh237 B
- 参考文档perks/validate/src/contracts.json358 B
- 参考文档perks/validate/test/case.json359 B
- 参考文档perks/validate/test/fixture/good.blueprint.json846 B
03 SKILL.md 正文
backbone — prove the bone before writing the flesh
The L++ preflight gate for spec-first development: a blueprint (the bone) must pass
lpp validate → TLC (bounded model check, EMPIRICAL) → TLAPS (deductive proof,
AXIOMATIC) before its compute units (the flesh) are implemented. This is Emerald §1.1b
made a governed pathway: intelligence at the edges, execution pre-proven.
This is the standard pattern for all future development:
- Spec as blueprints. Each component is an L++ blueprint; each decision table is a
separate ruleset JSON (law-as-data, changed by PR); each test is itself a blueprint
that DISPATCHes the component as a compute unit. Exemplar:
crucible/docs/spec/shapes/. - Backbone preflight. Run
preflightover the blueprint dir — every blueprint must pass all three layers (all-green or fail; empty discovery is a FAIL). - Only then implement the compute units the blueprints name, accepted by the test blueprints' fixtures.
Perks
validate— layer 1: L++ schema/load validation.TARGET_BLUEPRINT.tlc— layer 2: TLA+ generation + TLC.TARGET_BLUEPRINT(+TLC_TIMEOUT, default 120s).tlaps— layer 3: TLAPS proof generation + tlapm discharge.TARGET_BLUEPRINT(+TLAPM_TIMEOUT, default 240s).preflight— the full matrix overTARGET_DIR, recursively; per-file report + summary.
What to look out for
Each tool writes structured JSON (backbone_<perk>.json) into the record store — that
report is the certificate. Fail-closed everywhere: a missing prover binary is
status=fail (missing_tool), never a silent skip; an empty preflight discovery is a fail.
TLC/TLAPS here prove the state skeleton (gate NULL-safety, transition well-formedness,
TypeInvariant induction) with values abstracted — data-level safety_invariants are the
compute units' contracts, enforced by the test blueprints at implementation time.
Blueprint-authoring idioms the solvers enforce
- Gates NULL-safe in conjunction-first form (
x is not None and …) or gate on a scalar the compute outputs — the TLA generator inits context to NULL and branches disjunctions. - No comprehensions in gate expressions (untranslatable) — per-element checks are the compute unit's contract; the gate reads a scalar.
- Never name a transition
t_next— the generated lemma collides with theNextinduction lemma in TLAPS.
Requirements
python3 with the lpp package importable; tlc (+ java) for layer 2; tlapm for
layer 3. Nodes lacking a layer's prover refuse that perk (fail-closed) rather than pass.
04 同仓库更多 Skills
Vercel Deploy
暂无简介
Audio Transcribe
暂无简介
Speech (OpenAI TTS)
暂无简介
Skill Installer
暂无简介
demo-skill
A demo skill used as a test fixture for the validate perk. Use when checking that quick_validate.py accepts a well-formed SKILL.md.
demo-skill
A demo skill used as a test fixture for generating agents/openai.yaml. Use when validating the gen_openai_yaml perk.