roadmap: track Agent Zero milestones

[RedStar071]

Purpose

Track milestone-level progress for the Agent Zero roadmap and provide a central place to discuss changes to project direction.

The detailed roadmap is being introduced in ROADMAP.md and is organized around the project's find, fix, and verify principle.

Milestones

v0.1 — Resolve review feedback

• feat: v0.1 resolve review feedback with evidence #8
• ingest GitHub PR review feedback
• validate findings before acting
• apply narrowly scoped fixes
• verify with repository-native checks
• publish evidence and deterministic outcomes
• complete isolated production runner support

v0.2 — Proactive review and autofix

• feat: v0.2 add proactive review and confidence-gated autofix #9
• proactive PR diff analysis
• severity/confidence/evidence-based findings
• confidence-gated autofix
• repository-level autofix policy

v0.3 — Control plane and runner pools

• feat: v0.3 productionize control plane and runner pools #10
• production control plane
• task/event history
• model cost and usage tracking
• provider-neutral remote sandboxes
• runner pools and scheduling
• dashboard

v0.4 — Issue-to-PR workflows

• feat: v0.4 add bounded issue-to-PR workflows #11
• scoped issue ingestion
• repository investigation and acceptance criteria
• isolated implementation branches
• verification before PR creation
• human approval gates for high-impact work

v0.5 — Multi-provider source control

• feat: v0.5 add multi-provider source control support #14
• evolve the GitHub-specific integration boundary into provider-neutral contracts
• support GitHub / GitHub Enterprise, GitLab / GitLab Self-Managed, Bitbucket Cloud / Data Center, and Gitea / Forgejo where practical
• normalize pull requests, merge requests, reviews, approvals, checks/pipelines, issues, branches, and webhook events
• support mixed-provider deployments without leaking provider-specific semantics into the agent runtime
• add adapter capability detection and conformance tests

v0.6 — Configurable skills and MCP integrations

• feat: v0.6 add configurable skills and MCP integrations #15
• make Agent Skills discoverable and configurable per repository, task type, and agent role
• support explicit skill allow/deny policy and capability metadata
• provide curated, preconfigured MCP integration presets
• allow users and deployments to add local or remote MCP servers declaratively
• expose only authorized MCP tools through a provider-neutral tool registry
• add capability discovery, health diagnostics, trust boundaries, audit evidence, and granular tool policy
• keep credentials out of repository configuration, task logs, and model context unless explicitly required and safely scoped

Tracking policy

This issue tracks roadmap-level status and scope. Concrete implementation work should be split into focused issues and pull requests and linked back here.

Roadmap changes should preserve the core architectural constraints: provider-neutral agent runtime, execution behind the runner boundary, explicit repository policy, and verification before claiming success.