AI made the change. Show what was checked.
Replication Pack Studio turns one AI-assisted code change into an inspectable local evidence pack: target, selected evidence, command log, checksums, HTML report, deterministic ZIP export, and detached signature verification.
Founding design-partner price: $149 for one guided AI-assisted code-change handoff pack while the workflow is being validated.
Please do not send secrets, credentials, private keys, customer data, proprietary code, or sensitive transcripts in an initial message.
Current status: v0.1 technical preview, CLI-first, local-first, source/GitHub/developer-first. The pilot is for people who need a better handoff artifact around one real AI-assisted code change.
The gap is not code generation. The gap is proof of work.
AI-assisted development can produce a patch quickly. Review still needs context: what the change was supposed to do, what evidence was selected, which commands ran, what changed after testing, and what still needs human judgment.
Pilot question
Would you pay for a structured handoff pack that makes one AI-assisted change easier to review, explain, deliver, or revisit later?
Write down the charter, success criteria, and out-of-scope boundaries before the work becomes a pile of diffs and logs.
Collect selected inputs, outputs, command events, before/after test output, review notes, and supporting files in a folder-style .repack.
Run local checks over structure, references, checksums, report coverage, schemas, and obvious secret-pattern hits.
A real public-safe demo pack already verifies locally.
The current demo is synthetic and intentionally bounded: a toy percent-change helper, before/after tests, reviewed diff, command log, report, export ZIP, signature sidecar, and public key. It exists to show the handoff workflow without exposing private customer code. The folder-style .repack verifies locally; the exported ZIP is the artifact with detached signature verification.
Demo artifact
examples/replication-pack-ai-code-change-v0.1/example.repack
example.repack/
target-charter.md
manifest.json
decision-log.jsonl
evidence/commands.jsonl
evidence/inputs/
before-calculator.py
bug-report-and-ai-request.md
percent-change-tests.py
evidence/outputs/
after-calculator.py
before-test-output.txt
after-test-output.txt
code-change.diff
ai-assisted-change-summary.md
review-before-sharing-note.md
checksums.txt
verification-report.html
schemas/Verification Report: AI-assisted toy code-change review demo
Status: PASS
Verification checks:
manifest.json: ok
Manifest references: 22 ok, 0 missing
Decision log events: 1
Command events: 3
Checksum files: 22
Secret scan hits: 0
Schemas parsed: 2
The report is meant to be the first thing a reviewer opens.What a reviewer gets
- A short target/charter describing the intended change and success criteria.
- Selected input and output evidence, including before/after test output where applicable.
- A command log and exit-code summary instead of an unsupported “it worked” claim.
- Checksums for pack files and a generated human-readable HTML report.
- A deterministic ZIP export plus detached signature verification instructions when a shareable handoff artifact is needed.
Claim boundary
Replication Pack Studio shows what was packaged and checked. It checks pack structure, evidence/reference integrity, JSONL logs, report coverage, checksums, schema parseability, obvious secret-pattern hits, deterministic exports, and detached signature integrity for a supplied public key.
Verification and signing are limited evidence checks only. They do not prove truth, correctness, legal clearance, compliance, security, privacy, automatic redaction, customer suitability, completeness, or bug-free status. Human review still matters.
Good first design partners
The pilot is narrow by design. It is for one real handoff pain, not a broad platform promise.
You use Codex, Claude Code, Cursor, Copilot, or another AI-assisted workflow and want one change to be easier to review later.
You need to show a client not just the diff, but the evidence, commands, review notes, and limitations around a delivered change.
You are receiving AI-assisted PRs and want contributors to hand over evidence in a repeatable, inspectable format.
Your team is using coding agents and wants a stronger handoff artifact for review, audit, or onboarding.
You expect to revisit the decision later and need a compact archive of what was attempted and checked.
Regulated-use proof, formal security or compliance attestations, hands-free privacy scrubbing, legal clearance, or replacing human code review.
Design-partner pilot: one AI-assisted code-change handoff pack.
Founding price: $149. Bring one small AI-assisted change where a reviewer, client, maintainer, or future you would benefit from a structured handoff. Walter Claw Software will help package the target, selected evidence, command history, checksums, report, and review boundary, then use your feedback to decide what should become productized.
Please do not send secrets, credentials, private keys, customer data, proprietary code, or sensitive transcripts in an initial message.