docs: establish AI-native operating system

[kavatana]

Closes #5

Trace

• Accountable human: @kavatana
• Risk tier: R0
• Branch: docs/5-ai-native-operating-system
• Requested reviewer: @chamroeunhongleng

Outcome

Establish one honest, agent-readable operating system for CHNAI LAB work from member onboarding through issue, branch, draft PR, evidence, human review, merge, access review, and public or private AI-assistant delivery.

This moves organization defaults into GitHub's supported inheritance paths, makes GitHub the engineering system of record, and defines how the connected digital city can become more convenient without giving public agents private authority.

Scope And Non-Goals

Changed:

• Added governance, repository adoption, onboarding, and detailed human-agent workflow contracts.
• Added docs/AI_CONCIERGE_STANDARD.md for the portfolio, seven product specialists, authenticated assistants, Studio OS operator, and future agent gateway.
• Defined L0-L4 release levels, evidence classes, answer contracts, UX and security gates, product-specific claim boundaries, evaluation requirements, and MCP/A2A usage boundaries.
• Defined Slack, Gmail, LinkedIn, and X channel-adapter boundaries, including official OAuth-only authorization and human review for outbound claims or commitments.
• Added ten Mermaid diagrams across delivery, traceability, access, repository adoption, public/private agent routing, request lifecycle, and release state.
• Moved issue forms to .github/ISSUE_TEMPLATE/ and added decision, risk, owner, evidence, and verification fields.
• Added owner CODEOWNERS, Copilot entry instructions, Actions Dependabot, least-privilege CI, immutable Action pins, and branch-diff whitespace validation.
• Strengthened the verifier with file, policy, secret-marker, personal-email, commit-identity, inheritance-path, Mermaid, issue-form, CI, and concierge-contract checks.
• Updated the public profile, contributing guide, PR template, and security policy.

Intentionally unchanged:

• No product repository source or runtime behavior.
• No model, retrieval service, agent endpoint, MCP server, A2A Agent Card, LinkedIn/X connection, or action-capable agent is presented as live.
• No membership, team, repository access, visibility, billing, secrets, or production setting.
• No claim that private-repository branch protection is enforced on GitHub Free.

Evidence

Automated checks:

python3 scripts/verify_profile.py -> pass
PYTHONPYCACHEPREFIX=/tmp/chnai-profile-pycache python3 -m py_compile scripts/verify_profile.py -> pass
git diff --check -> pass
Markdown relative-link validation across repository Markdown -> pass
Mermaid CLI parse and PNG render for all three new diagrams -> pass
HEAD commit identity privacy check -> pass through repository verifier

Human manual verification still required:

• Review every rendered Mermaid diagram on GitHub.
• Open each inherited issue form and confirm field layout.
• Confirm the profile and policy links resolve after merge.
• Review the governance and concierge language, especially the R2/R3 boundary.
• Confirm that the portfolio is the only concierge described as LIVE.

AI Involvement

• Agent or tool: OpenAI Codex
• Agent contribution: repository audit, official-path verification, workflow and agent-city design, implementation, deterministic checks, Mermaid rendering, and draft PR preparation.
• Human decisions: @kavatana defined the AI-native collaboration goal, connected digital-city direction, team context, professional tone, traceability requirement, and human-curiosity boundary.
• Sensitive context shared with tools: none.

Risk And Rollback

• Main failure mode: a default guide becomes too heavy, unclear, inconsistent with a stricter product contract, or implies a capability is live before deployment.
• Mitigation: local repository rules retain precedence and can be stricter; the workflow defines explicit stop conditions, adoption states, release levels, evidence classes, and claim gates.
• Rollback: revert the relevant documentation commit; no product runtime or privileged setting changed.
• Remaining uncertainty: current private product repositories need separate adoption PRs, any live model-backed assistant will require R2 evidence and domain review, and technical branch enforcement requires a GitHub Team plan.

Public And Private Boundary

• No secrets, credentials, recovery codes, personal email addresses, private user data, or private strategy are included.
• The public portfolio concierge is explicitly separated from Studio OS, private tools, production, payments, and trading.
• No unsupported traction, revenue, impact, safety, compliance, certification, autonomy, security, or deployment claim is added.
• AI assistance is disclosed without fabricated human credit or evidence.
• Privileged organization settings remain unchanged.

Coordination

Draft PR #4 overlaps with attribution and starter-link guidance. This branch includes those corrections. If #4 merges first, this branch should be rebased before review; if this PR supersedes #4, close #4 with that reason instead of merging duplicate text.