---
schema_version: "newruntime-agent-readable-v0.2"
type: "project_build_brief"
stable_id: "project_build_brief:verifiable-ai-coding-workflow"
project_slug: "verifiable-ai-coding-workflow"
project_title: "Verifiable AI Coding Workflow"
buildability_class: "B2"
verification_status: "locally-reproduced"
proof_artifact_count: 6
buildability_axis_count: 7
replay_evidence_count: 2
project_status: "candidate"
last_verified: "2026-08-06"
record_date: "2026-08-06"
date_kind: "last_verified"
routes: {"html":"https://newruntime.com/projects/verifiable-ai-coding-workflow/","project_markdown":"https://newruntime.com/projects/verifiable-ai-coding-workflow.md","project_json":"https://newruntime.com/projects/verifiable-ai-coding-workflow.json","acceptance":"https://newruntime.com/projects/verifiable-ai-coding-workflow/acceptance.md"}
---

# Build brief: Verifiable AI Coding Workflow

## Desired outcome

Turn an implementation idea into a bounded coding task with checks and reviewable evidence.

## Wow moment

The agent does not just produce a patch; it leaves commands, changed files, acceptance checks, and a human-readable handoff.

## Known inputs

- One bounded implementation idea
- Repository with runnable checks
- Files or modules likely in scope
- User acceptance criteria and explicit non-goals

## Available tools

- Repo search
- Shell
- Test runner
- Git diff

## Proof status

- Verification status: locally-reproduced.
- Last verified: 2026-08-06.
- Expected artifacts: 3.
- Provider scopes: 3.

## Fixture outputs

- Project JSON contract: dist/projects/verifiable-ai-coding-workflow.json (https://newruntime.com/projects/verifiable-ai-coding-workflow.json). Check: The JSON route exposes proof, handoff, permissions, acceptance checks, and source URLs in one replayable contract.
- Build brief handoff: dist/projects/verifiable-ai-coding-workflow/build-brief.md (https://newruntime.com/projects/verifiable-ai-coding-workflow/build-brief.md). Check: The build brief preserves scope, non-goals, build steps, questions, and approval gates without requiring the original chat.
- Acceptance checklist: dist/projects/verifiable-ai-coding-workflow/acceptance.md (https://newruntime.com/projects/verifiable-ai-coding-workflow/acceptance.md). Check: The checklist separates required checks, permission scopes, reproduction commands, known limits, and failure conditions.
- Clean-room replay fixture: scripts/project-proof-depth-v2-fixture-test.mjs. Check: The fixture replays a bounded coding task in a temporary directory, proves the initial failure and final pass, and verifies that an unrelated file is unchanged.
- Independent replay kit: scripts/project-proof-independent-replay-kit.mjs. Check: The kit starts failing, exposes no ready solution, protects every out-of-scope file by checksum, and refuses to emit a receipt without a non-author attestation and a passing final check.
- Independent agent replay receipt: src/data/project-proof-receipts/verifiable-ai-coding-workflow-independent-agent-v1.json (https://newruntime.com/projects/verifiable-ai-coding-workflow.json). Check: A separately operated Codex received only the isolated kit, observed the initial failure, changed only src/add.mjs, passed the final check, and produced an attestation that the primary verifier accepted.

## Reproduction commands

1. npm run project-proof:replay:test
2. npm run project-proof:replay:prepare
3. npm run validate:content
4. npm run build
5. npm run validate:agent-indexes

## Expected artifacts

- A task brief with scope, acceptance checks, explicit non-goals, and files or modules likely in scope.
- A focused diff plus command evidence for at least one real check or a reproducible blocker.
- A final handoff that separates implemented changes, verification evidence, residual risk, and approval-gated actions.

## Buildability breakdown

Assessment level: non-author-replayed. Assessed at: 2026-08-06.

| Axis | Level | Evidence | Bottleneck |
| --- | --- | --- | --- |
| time | high | The clean-room task reaches a checked result in one bounded local run, while the public blueprint targets first value within one to two hours. | Repository setup and test availability still determine the real start time. |
| code burden | medium | The workflow needs a developer-readable task and a focused source edit, but no bespoke application or new service. | A nontechnical owner still needs an agent or engineer to interpret failures. |
| integration burden | high | The replay uses only repository files, Node, a test runner, and the existing agent handoff contract. | Repositories without a runnable local check need an integration step before replay. |
| operational burden | high | The first proof is local and temporary, requires no server, and deletes no shared or production state. | Long-running or deployment tasks would need additional recovery and monitoring rules. |
| permission clarity | high | Read, local-write, commit, push, deploy, and external-write boundaries are named separately in the build brief and acceptance contract. | A broader task can still create ambiguity unless exact files and approvals are fixed before editing. |
| reproducibility | high | A separately operated Codex received only the isolated public kit, reproduced the initial failure, changed the one allowed file, and passed the same acceptance check without inspecting the author solution. | A real external repository replay would test integration burden beyond this bounded fixture. |
| failure recovery | high | The replay records the initial failing check, preserves an unrelated file, performs no remote action, and leaves explicit residual-risk and approval sections. | Production rollback and credential recovery are outside this local proof. |

## Replay evidence

### clean-room-bounded-edit-v1

- Date: 2026-08-06.
- Kind: automated-clean-room.
- Instructions: public-build-brief-and-acceptance.
- Environment: Temporary local directory with a bounded Node fixture and no network or remote writes.
- Outcome: passed.

Artifacts:

- Task brief: task-brief.json. Check: Names the allowed source file, acceptance command, non-goals, and approval-gated actions.
- Focused implementation: src/add.mjs. Check: Changes only the named implementation file while the unrelated note retains its original checksum.
- Real test: test/add.test.mjs. Check: Fails against the initial implementation and passes after the focused edit.
- Structured handoff: handoff.json. Check: Separates implemented files, verification evidence, residual risk, and approval-gated actions.

Limitations:

- This replay proves the deterministic local contract and failure boundary, not a real external repository integration.
- No human or separately operated agent independently interpreted the brief in this first replay.

### independent-codex-replay-v1

- Date: 2026-08-06.
- Kind: independent-agent.
- Instructions: public-build-brief-and-acceptance.
- Environment: Isolated temporary replay kit; the executor received no parent conversation or main-repository context and used no network or remote writes.
- Outcome: passed.

Artifacts:

- Independent attestation: reviewer-attestation.json. Check: Names an independent agent, records initial failure and final pass, lists only src/add.mjs as changed, and attests that the author solution was not inspected.
- Verified replay receipt: src/data/project-proof-receipts/verifiable-ai-coding-workflow-independent-agent-v1.json (https://newruntime.com/projects/verifiable-ai-coding-workflow.json). Check: The primary verifier checked protected-file hashes, rejected unexpected files, reran the acceptance test, and recorded zero unexpected files.

Limitations:

- This proves a bounded non-author replay of the public workflow contract, not integration into a production repository.
- Reviewer identity is a local agent attestation and is not cryptographically verified.

## Provider scopes

| Provider | Access | Minimum scope | Approval gate |
| --- | --- | --- | --- |
| Local repository | write | Focused source, test, or documentation files inside the shared worktree. | Human names the task and repository before edits start. |
| Git remote | publish | Focused commit and push only when explicitly requested. | Commit, push, deploy, and external writes require a separate explicit owner command. |
| Test runner or local dev server | read | Named checks and local rendered routes needed for the accepted task. | Human or repo docs identify which checks count for acceptance. |

## Privacy constraints

- Data touched: Repository source, tests, local command output.
- Use only publishable public source URLs in public output.
- Do not expose private discovery feeds, raw exports, credentials, drafts, or internal handoff artifacts.

## Required approvals

- Local file edits only; commits and deploys remain explicit approvals
- read: Human names the task and repository before inspection starts.
- write: Human task request authorizes local edits; commit/push/deploy still need explicit approval.

## Build steps

1. Convert the idea into a task brief with scope, non-goals, and acceptance checks.
2. Let the agent inspect files and report the intended edit path before changing code.
3. Apply a focused diff and keep unrelated dirty work out of the change.
4. Run the named checks and capture command evidence, failures, and residual risk.
5. Produce a handoff that names changed files, checks run, and any approval still required.

## Acceptance checks

- The task brief can be read without the original chat and still preserves scope and non-goals.
- The final diff is focused on the named files or the agent explains every extra touched file.
- At least one real check command runs, or the blocker is explicit and reproducible.
- The handoff separates implemented changes, verification evidence, unresolved risks, and approval-gated actions.

## Explicit non-goals

- Do not commit, push, deploy, or call external write APIs without a separate explicit approval.
- Do not broaden the task after inspection unless the human accepts the expanded scope.
- Do not treat model output as verification evidence without a command, rendered check, or direct artifact.

## Questions before external systems

- Which repository and branch/worktree should be used?
- Which checks are required for acceptance?
- Are commit, push, deploy, or external writes explicitly in scope?

## Related routes

- html: https://newruntime.com/projects/verifiable-ai-coding-workflow/
- project_markdown: https://newruntime.com/projects/verifiable-ai-coding-workflow.md
- project_json: https://newruntime.com/projects/verifiable-ai-coding-workflow.json
- acceptance: https://newruntime.com/projects/verifiable-ai-coding-workflow/acceptance.md