Define Phase 0 architecture and counterparty onboarding

[chefmrfrizzle]

What changed

• adds the missing protocol architecture baseline
• defines canonical terminology and 20 protocol invariants
• records accepted, proposed, deferred, and blocked decisions
• adds a 24-test acceptance plan and prioritized risk register
• designs sandbox-first onboarding for every counterparty
• proposes an evidence-driven innovation backlog
• captures desktop/mobile UX audit evidence
• records the work as LearningEpisode 0001
• documents the verified GitHub security baseline

Why

The repository contract requires Phase 0 architecture and decision work before production implementation. This converts the original concept into explicit build boundaries, review gates, and testable claims.

Impact

No production code, payment rail, token, restricted workload, or scientific result is introduced. Signature suite, first benchmark acceptance policy, reproduction independence, and production retention remain blocked pending evidence and review.

Validation

• repository baseline passed locally
• Markdown links resolved
• workflow and Dependabot YAML parsed
• credential and private-key pattern scan passed
• desktop and mobile prototype states captured and inspected
• public main protection and live security controls verified

[Copilot]

Pull request overview

Defines the Phase 0 architecture, governance boundaries, onboarding model, and repository safeguards.

Changes:

• Establishes canonical terminology, invariants, decisions, risks, and acceptance gates.
• Documents onboarding, UX evidence, innovation priorities, and learning episodes.
• Strengthens repository credential and full-diff validation.

Reviewed changes

Copilot reviewed 18 out of 20 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tests/test_repository_baseline.sh	Adds baseline regression tests.
scripts/check_repository_baseline.sh	Expands credential and diff checks.
docs/phase-0/TERMINOLOGY_AND_INVARIANTS.md	Defines protocol terms and invariants.
docs/phase-0/RISK_REGISTER.md	Records prioritized Phase 0 risks.
docs/phase-0/README.md	Indexes the decision package.
docs/phase-0/NEXT_BUILD_PROMPTS.md	Sequences gated follow-up work.
docs/phase-0/LEARNING_EPISODE-0002.md	Records readiness-review findings.
docs/phase-0/LEARNING_EPISODE-0001.md	Records Phase 0 bootstrap evidence.
docs/phase-0/INNOVATION_BACKLOG.md	Prioritizes evidence-driven ideas.
docs/phase-0/DECISION_REGISTER.md	Tracks decisions and ownership.
docs/phase-0/COUNTERPARTY_ONBOARDING.md	Designs sandbox-first onboarding.
docs/phase-0/audit/UX_AUDIT.md	Captures prototype UX findings.
docs/phase-0/ARCHITECTURE_REVIEW.md	Assesses architecture readiness.
docs/phase-0/ACCEPTANCE_TEST_PLAN.md	Defines 24 acceptance gates.
docs/ARCHITECTURE.md	Establishes the protocol architecture.
.github/workflows/repository-baseline.yml	Runs regressions and full-history checks.
.github/REPOSITORY_SETTINGS.md	Documents repository security controls.
.github/CODEOWNERS	Adds a second owner account.

Suppressed comments (2)

docs/ARCHITECTURE.md:80

• The first-slice acceptance sequence also omits the authorized WorkRequest, so the slice could satisfy this list while executing directly from a problem statement. Include the canonical authorization object before compiling the graph.

1. a valid ProblemPassport;
2. deterministic compilation to a WorkGraph;
3. one locally executable TaskSpec;

docs/phase-0/NEXT_BUILD_PROMPTS.md:23

• Prompt 5 skips the WorkRequest authorization/funding boundary defined by the canonical terminology and instead directs implementation straight from ProblemPassport to the adapter and graph. Following this prompt would bake the architecture inconsistency into the first executable slice.

> Implement the selected benchmark from ProblemPassport through TaskAdapter, WorkGraph, one bounded local worker, output commitment, and unsigned receipt structure. Pin and record the execution environment. Enforce resource/network/filesystem limits. Prove deterministic identifiers and replay safety. Do not call receipts cryptographically signed until the approved signature gate passes.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[11BUSD]

Security Review — Baseline Regression Tests (AR-003 + Nested Credential Paths)

Outcome

✅ AR-003 false-green risk is substantially closed
✅ Nested forbidden credential filename risk is substantially closed

Evidence observed

• Workflow now checks out full history (fetch-depth: 0).
• Workflow passes BASE_SHA and HEAD_SHA for PR-aware comparison.
• Baseline script now prefers git diff --check BASE...HEAD when both commits exist.
• Script falls back to recursive commit-tree checking for local/new-root scenarios.
• Basename scanning now iterates all tracked paths (git ls-files -z) and blocks forbidden basenames at any depth.
• Regression tests explicitly reproduce:
  • nested forbidden credential basenames
  • multi-commit diff whitespace regression that would have passed under HEAD-only checking

Security assessment

This is a meaningful fix with executable regressions. It upgrades from snapshot-style linting to change-range validation, which is the right control for PR correctness.

Remaining hardening opportunities (recommended)

1. Extend forbidden basename set (e.g., id_ecdsa, *.pem, *.p12, secrets.yml) with policy rationale.
2. Add symlink and case-variance tests (.ENV, Credentials.json) if cross-platform behavior matters.
3. Add one negative-control test showing allowed examples (.env.example, docs fixtures) remain valid.
4. Keep secret-pattern checks as defense-in-depth; avoid treating regex-only scans as complete secret detection.

Merge readiness for this security slice

Ready for merge from a baseline-check perspective, pending required human approval gates and conversation resolution.