szl-reproducibility-capsule
Creates an exact-byte manifest (path, size, SHA-256) for the explicitly selected files of an experiment, verifies them later as changed / missing / unsafe, and validates a bounded replay declaration (inputs, source, environment, plan, outputs) without executing anything. Use when handing an analysis to a collaborator, archiving a result, preparing a replication or when the user asks 'did anything change since the run'. Not a container, not a signature, and it does not reproduce the experiment.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-15
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- szl-reproducibility-capsule
- license
- Apache-2.0
- description
- Creates an exact-byte manifest (path, size, SHA-256) for the explicitly selected files of an experiment, verifies them later as changed / missing / unsafe, and validates a bounded replay declaration (inputs, source, environment, plan, outputs) without executing anything. Use when handing an analysis to a collaborator, archiving a result, preparing a replication or when the user asks 'did anything change since the run'. Not a container, not a signature, and it does not reproduce the experiment.
02 资源文件 11
- 资产文件assets/example.json184 B
- 资产文件assets/protocol.txt270 B
- 资产文件assets/replay-example.json1.1 KB
- 脚本assets/replay/analysis.py110 B
- 资产文件assets/replay/environment.json77 B
- 资产文件assets/replay/input.txt58 B
- 资产文件assets/replay/plan.json107 B
- 资产文件assets/replay/result.txt46 B
- 脚本kernel.py15.4 KB
- 参考文档references/contracts.md5.3 KB
- 脚本scripts/run.py3.4 KB
03 SKILL.md 正文
Reproducibility capsule
Pick the files that define an experiment and freeze their bytes. Later, prove they are the same bytes or list exactly which ones moved. Python 3.10+, stdlib, offline.
Use when
- Archiving the input tables, scripts, environment file and reference outputs behind a figure.
- A collaborator re-runs your pipeline and gets different numbers: verify the capsule first.
- Preparing a replay package for a separate sandbox review with declared seed and limits.
Quick start
python scripts/run.py assets/example.json --root .
Creates a szl.reproducibility-capsule.v1 record listing assets/protocol.txt (270 bytes, its
SHA-256), the synthetic metadata, signed: false and a capsule_sha256. Verification on retained
bytes reports changed, missing and unsafe or unreadable files separately. assets/replay-example.json
adds a replay declaration over the files in assets/replay.
Creating and verifying
szl_make_capsule(root, files, metadata=None, replay=None) accepts plain path lists or role-tagged
entries (path plus role). Paths must be canonical relative POSIX paths, unique, within hard
count and byte bounds; traversal, symlinks, junctions, reparse points and credential-like names
are refused. szl_verify_capsule(root, capsule) re-hashes. All bytes matching plus a complete,
valid replay declaration yields replay_ready: true, meaning prepared for a separate sandbox review;
execution: NOT_RUN, capability_denial: DECLARED_ONLY and tolerance_application: NOT_RUN stay
explicit. Read references/contracts.md before authoring a replay.
Keep an independently retained copy of the manifest when integrity across sessions matters: an
unsigned self-digest cannot detect replacement of the whole manifest. authentic: false stays false
even when every byte matches; real signing needs a key and a verifier (see szl-session-receipt for the
optional signing path).
What it does not do
Never executes argv or source, installs dependencies, launches processes, applies numerical comparisons to outputs, or inspects file contents for secrets (credential-name guards look at names). Limits: 128 files, 8 MiB per file, 32 MiB total. It cannot prove source safety, enforce a sandbox or verify preregistration. Hashes are not signatures; readiness is not reproduction.
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.