AGENT SKILL

szl-session-receipt

Writes a verifiable record of an analysis session: SHA-256 of every input, script and output file, the commands and environment as declared, a root digest, and a plain-language Methods paragraph; later re-hashes the files to report MATCH / MISMATCH / MISSING. Use at the end of a Claude Science or notebook session, when handing an analysis to a collaborator, when a journal asks for a reproducibility statement, or when the user says 'record what we just did'. Not a signature of 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-session-receipt
license
Apache-2.0
description
Writes a verifiable record of an analysis session: SHA-256 of every input, script and output file, the commands and environment as declared, a root digest, and a plain-language Methods paragraph; later re-hashes the files to report MATCH / MISMATCH / MISSING. Use at the end of a Claude Science or notebook session, when handing an analysis to a collaborator, when a journal asks for a reproducibility statement, or when the user says 'record what we just did'. Not a signature of correctness.

02 资源文件 7

03 SKILL.md 正文

Session receipt

A methods section that can be checked. While working, keep a manifest of the files read, the scripts run, the files produced and the commands used. record hashes all of it and writes the paragraph; verify re-hashes it tomorrow, or in a reviewer's hands, and says whether anything changed. Stdlib only, offline; signing is optional and absence is stated, never hidden.

Use when

  • Ending an agent-assisted analysis: "record this session" produces the receipt and the paragraph.
  • Submitting: the Methods paragraph and the receipt file go in supplementary material.
  • Receiving a colleague's analysis: run verify against their files before reading their conclusions.
  • Re-running months later: a MISMATCH on the raw data file explains the different numbers before anyone argues.

Quick start

python scripts/run.py record assets/example.json --root assets/project --output /tmp/receipt.json
python scripts/run.py verify /tmp/receipt.json --root assets/project
python scripts/run.py methods /tmp/receipt.json

The synthetic project gives "status": "RECORDED", verify gives "status": "MATCH" over 3 files, and the paragraph reads:

Analysis for SYNTHETIC dose-response reanalysis was performed by an AI agent (Claude (Claude Science)) between 2026-10-01T14:00:00Z and 2026-10-01T14:25:00Z. It read 1 input file(s), executed 1 script(s) via 1 recorded command(s), and produced 1 output file(s); the SHA-256 digest of every file is recorded in session receipt e042750035a157ae (root digest ...). The environment was declared as Python 3.12.1 with numpy 2.1.0, pandas 2.2.3. The receipt is unsigned; integrity can be re-verified by re-hashing the listed files, authenticity is not established.

Manifest

{"project": "...", "session_id": "...", "started_at": "...Z", "finished_at": "...Z",
 "actor": {"kind": "agent|human", "name": "..."},
 "inputs": ["data/raw.csv"], "code": ["analysis.py"], "outputs": ["results/summary.csv"],
 "commands": ["python analysis.py"],
 "environment": {"python": "3.12.1", "platform": "...", "packages": {"numpy": "2.1.0"}}}

Paths are relative to --root and may not escape it. Fields under declared_only (commands, environment, actor, timestamps) are recorded as stated, not observed. Optional --sign-key KEY.pem produces a DSSE signature when szl-receipt-dsse is installed; otherwise the receipt says UNSIGNED.

What it does not do

A MATCH means the files are byte-identical to the recorded ones. It does not mean the analysis is correct, that the recorded commands were the only ones run, or that the declared environment was real. It does not capture files you did not list. Details: references/contract.md.

04 同仓库更多 Skills

szl-holdings
szl-holdings/szl-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.

★ 0
szl-holdings
szl-holdings/szl-skills

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.

★ 0
szl-holdings
szl-holdings/szl-skills

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.

★ 0
szl-holdings
szl-holdings/szl-skills

demo

暂无简介

★ 0
szl-holdings
szl-holdings/szl-skills

demo

暂无简介

★ 0
szl-holdings
szl-holdings/szl-skills

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.

★ 0