Developer experience: CLI, privacy heartbeat, design partners, contribution model

[aniketh-maddipati]

What changed

This PR replaces the legacy security-assessment CLI with a new Typer-based AgentMint command surface built around instrumentation, verification, and operator ergonomics. The new entrypoint adds init, notarise, verify, export, plan, and chain, along with operational commands for doctor, show, privacy, watch, and actions. It also introduces shared CLI helpers for styling, config discovery, project scanning, and receipt rendering; adds provider and verification modules needed by the new workflow; updates the decorator path so the quickstart flow can emit receipts into the configured sink; and rewrites the top-level docs, examples, and CLI E2E coverage around that new experience.

Why this matters for users

Before this change, the repository still centered an older assessment-oriented CLI that did not match the current notary/runtime positioning described in the product direction. That made the first-run experience harder to understand, mixed two different product stories in one package, and left no end-to-end command flow for initializing a project, emitting receipts, inspecting them, checking privacy posture, and packaging evidence. The effect for users was extra onboarding friction and a mismatch between the library's public runtime direction and the CLI/documentation they actually encountered.

Root cause and fix

The root cause was that the repo had evolved its runtime and receipt model faster than its command surface, docs, and examples. This PR brings those pieces back into alignment by removing the legacy assessment stack, wiring the CLI around the current receipt lifecycle, and updating the repository narrative to focus on signed evidence, offline verification, privacy posture, and design-partner positioning. The fix also keeps generated local artifact directories out of the PR so the review stays scoped to source, docs, tests, and workflow changes.

Validation

I validated the rewritten flow with pip install -e ".[dev,cli]", agentmint --help, pytest, mypy --strict agentmint, and ruff check agentmint tests/cli. I also exercised the new CLI path in a temporary workspace: agentmint init --yes, a minimal decorated agent run that emitted a receipt, and follow-up agentmint verify, agentmint doctor, agentmint show, and agentmint privacy checks. The only review step that did not complete cleanly was vet, because the Codex agentic harness hit its input-size limit before analysis could finish.