szl-skill-conformance
Checks retained scientific skill-sharing evidence: exact local source and supplied host-export resource bytes, explicit import aliases, task and result digests, positive versus irrelevant-task trigger declarations, and optional artifact-to-trace byte bindings. Use when a shared skill appears imported but invocation is uncertain, when comparing retained trials across hosts, or when an update may invalidate an earlier result. Offline checking never executes the candidate and never certifies real host invocation or scientific correctness.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-15
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- szl-skill-conformance
- license
- Apache-2.0
- description
- Checks retained scientific skill-sharing evidence: exact local source and supplied host-export resource bytes, explicit import aliases, task and result digests, positive versus irrelevant-task trigger declarations, and optional artifact-to-trace byte bindings. Use when a shared skill appears imported but invocation is uncertain, when comparing retained trials across hosts, or when an update may invalidate an earlier result. Offline checking never executes the candidate and never certifies real host invocation or scientific correctness.
02 资源文件 5
- 脚本kernel.py17.8 KB
- 参考文档references/contract.md4.0 KB
- 参考文档references/use-record.md701 B
- 脚本scripts/run.py901 B
- 脚本scripts/use_record.py896 B
03 SKILL.md 正文
Skill conformance
Use this when asked whether an imported scientific skill actually worked. Separate local source integrity, supplied export binding, declared trial consistency, actual host observation, and scientific result validity. Never combine them into "verified".
-
Ask for the exact source skill directory, the host-exported resource directory, and retained task/result files. An import screenshot or catalog entry is not an invocation trace. Do not fabricate absent files or execute unknown candidate code.
-
Read the evidence contract. Record the source revision separately. This checker hashes local bytes, not Git publication or signatures.
-
Run the reviewed checker with Python 3.10+ and
-B(no dependencies, network, credentials, provider writes, model weights, or subprocess execution):python -B scripts/run.py --source PATH --host-export EXPORT --trials trials.json --trial-root RETAINEDWithout an export or trials, report
INCOMPLETE. Do not turn missing evidence into a successful empty comparison.--imported-name ALIASdeclares a display-name mapping; it does not excuse modified payload bytes. -
Include a task that should invoke the skill and an unrelated task that should not. Bind both to retained results. Inspect whether the controls are meaningful: this checker cannot judge task relevance or output correctness from hashes.
-
Report
actual_host_invocation: NOT_VERIFIEDeven when supplied records are internally consistent. Synthetic fixtures and operator exports cannot prove the app executed a skill. Unknown evidence fields, includingverified, fail closed. -
For a real Claude Science pilot, collect evidence inside the actual app through documented interfaces, retain exports, and have a researcher inspect the task, selected skill, resources, result and irrelevant-task control. Do not inject app state, change private databases, bypass permissions or auto-update imports.
For a retained result artifact, the optional skill-use record contract
binds its bytes and an optional trace file to the exact local skill bundle. Run
python -B scripts/use_record.py --source SKILL_DIRECTORY --record use-record.json --retained RETAINED_DIRECTORY.
An absent trace is UNKNOWN; even a matching supplied trace is NOT_VERIFIED as
actual host invocation. The checker does not authenticate the declared Git revision,
trace origin, or scientific result.
This source candidate is outside the immutable v0.4.0 and v0.5.0-rc.3 releases. No claims of cross-host portability, agent efficacy, novelty or scientific validation follow from its offline tests. Review every resource before importing this skill.
04 同仓库更多 Skills
szl-unit-invariants
Checks declared SI dimensions, allowlisted unit conversions, expected ranges and fixed pointwise invariants (equal, sum, difference, product, ratio) over selected measurement or simulation columns, flagging incompatible quantities, scale errors of 1000x and broken conservation checks. Use when merging instrument exports with different units, validating simulation output against conservation laws, or when a reviewer asks how unit errors were ruled out. Not an uncertainty-propagation or symbolic tool.
szl-uncertainty-lineage
Propagate declared measurement uncertainty through a small arithmetic dependency graph using a first-order delta approximation. Use when a researcher needs to see how corrected inputs or shared-source correlation change a reported standard uncertainty; require unit checks separately.
szl-typesafe-ai
TypeSafe/Jev: Choice/Noul/Score in code, not TypeScript types. System One models turn natural language and application state into probabilities that software can compose. Use when building routing, ranking, extraction, verification, or replacing an LLM prompt-and-parse step with structured decisions; also when brainstorming what TypeSafe could make possible. Do not use for TypeScript/Java/Rust static types, Zod/Pydantic/mypy/JSON Schema/protobuf codegen, or asking whether a pipeline is type-safe. At SZL, Jev is an optional second reader only.
demo
暂无简介
demo
暂无简介
szl-skill-update-review
Compare pinned local skill packages using inventories and a retained lock; report changed skills, files, declarations and reruns without approving safety or validity.