Roadmap¶
The plan lives in roadmap/roadmap.yaml and in the repository issues: milestones, epics and
tasks, each task linked to its issue and, when it has one, to its specification. GitHub holds
the state of every issue; scripts/roadmap.py sync copies it into the file, and this page is
generated from the file at every build of the site.
Execution comes first: it is the part with the most proven practice behind it. The expert council is the most visible part but the least proven one, and it is built when the rest of the loop can receive its output.
| Milestone | Gate | Release | Done | Open |
|---|---|---|---|---|
| M0 · Foundations | wayloop init on an empty repository produces a structure that passes wayloop validate. | 0.0.1 | 56 | 21 |
| M1 · Engine core | The engine replays a ledger to the same state on every platform and drives the issue cycle in a dry run on local fixtures. | 0.1.0 | 6 | 19 |
| M2 · Tracker and loop | Wayloop runs the issues of its own repository end to end: claim, build, verify, pull request, with gates on GitHub. | 0.2.0 | 0 | 41 |
| M3 · Living specifications and survey | A change on an existing codebase is surveyed, specified as a delta, implemented and merged into the living specification. | 0.3.0 | 0 | 15 |
| M4 · Upstream planning | A new project goes from a brief to a council report, a roadmap, user stories and generated issues, with every gate approved by a person. | 0.4.0 | 0 | 14 |
| M5 · Adapters and distribution | Wayloop installs from crates.io, PyPI and npm, and as a plugin for Claude Code and Codex, in an external repository in less than five minutes on Linux, macOS and Windows. | 0.5.0 | 1 | 23 |
| M6 · Benchmark and hardening | The benchmark results are published, the threat model is tested, performance budgets hold and no blocking defect is open. | 0.9.0 | 0 | 12 |
| M7 · Version 1.0 | Wayloop 1.0.0 is published on every channel with frozen contracts and a stability policy. | 1.0.0 | 0 | 4 |
M0 · Foundations¶
Gate: wayloop init on an empty repository produces a structure that passes wayloop validate.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| Repository and governance | Enable GitHub Pages and upload the social preview | #1 | Open | |
| Repository and governance | Cap the Pages deploy timeout at the allowed maximum | #2 | Done | |
| Repository and governance | Specs, roadmap as code, sprints and process checks | #15 | Done | |
| Repository and governance | Decide on ADR 0006 and ADR 0007 | #10 | Done | |
| Repository and governance | Header colour of the site follows the brand palette | #9 | Done | |
| Repository and governance | Roadmap to version 1.0 with specifications, ADRs and issues | #43 | Done | |
| Repository and governance | README and documentation follow the structure of the shared project template | #44 | Done | |
| Repository and governance | Plan the work checkout, the pull request ledger, naming rules in the engine, token saving and agent compatibility | #157 | Done | |
| Repository and governance | Plan the process policy in the engine and the loop improvements | #171 | Done | |
| Repository and governance | Plan the process and system skills, agents and hooks | #191 | Done | |
| Repository and governance | Plan one contract on every channel with translation adapters and a channel reconciler | #220 | Done | |
| Repository and governance | Plan the execution modes, loop continuation, batches and composed agents | #242 | Done | |
| Process parity with the project template | Process scripts are covered by tests that run in CI | #137 | Done | |
| Process parity with the project template | Issue forms for task, bug and specification request | #138 | Done | |
| Process parity with the project template | Issue titles state the outcome, are unique and carry no type prefix | #45 | Done | |
| Process parity with the project template | Roadmap file reproduced on GitHub by a script in the repository | #139 | Done | |
| Process parity with the project template | Required status checks kept in step with the workflows | #140 | Done | |
| Process parity with the project template | Workflows linted, actions pinned and dependencies scanned | #141 | Done | |
| Process parity with the project template | Weekly maintenance workflow | #142 | Done | |
| Process parity with the project template | Claude Code hooks for documentation impact and formatting | #143 | Done | |
| Process parity with the project template | Documentation kept in sync with the project by a deterministic check | #144 | Done | |
| Process parity with the project template | Automated review of pull requests and documentation | #145 | Done | |
| Process parity with the project template | Decide on a GitHub Projects board for the roadmap | #146 | Open | |
| Process parity with the project template | Traceability rules of the project template written down and checked | #158 | Done | |
| Process parity with the project template | Session briefing proposes only issues that can start now | #224 | Done | |
| Process parity with the project template | Defects found during any work are fixed in the same change or filed at once | #226 | Done | |
| Process parity with the project template | Quality problems surface and are fixed without the maintainer watching | #240 | Done | |
| Process parity with the project template | Roadmap sync tolerates issues created moments before their roadmap change | #249 | Done | |
| Process parity with the project template | Roadmap sync runs only when a real difference can appear | #280 | Done | |
| Process parity with the project template | Pull requests fail only on their own changes, not on a new toolchain | #283 | Done | |
| Process parity with the project template | Every dependency of every kind is checked against one licence allow-list | #286 | Done | |
| Process parity with the project template | A documentation waiver covers only a branch whose every commit carries it | #289 | Done | |
| Process parity with the project template | Branch names and closing keywords have one definition used by every script | #297 | Done | |
| Process parity with the project template | Every rule and value has one source and duplicated code fails the build | #298 | Done | |
| Process parity with the project template | The pure crates keep no duplicated helper after the contract generator | #299 | Done | |
| Artifact model | Approved specification of the artifact model and schemas | #3 | 0001 | Done |
| Artifact model | Decide on ADR 0008 on identifiers and traceability | #4 | Done | |
| Artifact model | wayloop init creates a project structure that validates | #5 | 0001 | Open |
| Artifact model | Artifacts and their links are validated | #6 | 0001 | Open |
| Artifact model | Naming and linking rules validated by the engine | #160 | 0001 | Open |
| Artifact model | Existing specifications follow the Markdown rules of the artifact model | #229 | 0001 | Open |
| Artifact model | Complete the error report of ADR 0009 in the core | #252 | 0001 | Done |
| Artifact model | Add canonical JSON, SHA-256 and the envelope to the core | #253 | 0001 | Done |
| Artifact model | Refuse in CI any change to a released schema snapshot | #254 | 0001 | Done |
| Artifact model | Interpret the JSON Schema subset in the core | #255 | 0001 | Done |
| Artifact model | Validate a brief through the entry point and wayloop validate --input | #256 | 0001 | Done |
| Artifact model | Load repository files into the canonical input of validate | #257 | 0001 | Open |
| Artifact model | Validate milestones, epics, stories, tasks and gate issues | #258 | 0001 | Open |
| Artifact model | Validate council reports, living specifications and change deltas in normalised form | #259 | 0001 | Open |
| Artifact model | Validate hand-over records and hand-over inputs | #260 | 0001 | Open |
| Artifact model | Validate the shape of pull request and commit inputs | #261 | 0001 | Open |
| Artifact model | Check identifier forms and uniqueness across the set | #262 | 0001 | Open |
| Artifact model | Check links and the hierarchy between artifacts | #263 | 0001 | Open |
| Artifact model | Check parts of artifacts and change deltas | #264 | 0001 | Open |
| Artifact model | Normalise living specifications from Markdown | #265 | 0001 | Open |
| Artifact model | Check epic and task titles | #266 | 0001 | Open |
| Artifact model | Validate and produce branch names and commit headers | #267 | 0001 | Open |
| Artifact model | Run wayloop validate on this repository in CI | #268 | 0001 | Open |
| Artifact model | Document the artifact model | #269 | 0001 | Open |
| Artifact model | Allocate the next identifier of a kind | #270 | 0001 | Open |
| Artifact model | Plan and tasks of the artifact model with their issues | #276 | 0001 | Done |
| Artifact model | Contracts are generated from one source and versioned like Fairmark | #291 | Done | |
| Artifact model | Contract types derive their JSON Schema within the interpreter subset | #293 | Done | |
| Artifact model | Schema snapshots are generated and their history is kept per contract | #294 | Done | |
| Artifact model | The envelope and the error report are generated from their types | #295 | Done | |
| Artifact model | The map of lines is typed in the contract of validate | #307 | Open | |
| Agent process | Claude Code and Codex mirrors generated from .agents | #7 | Done | |
| Agent process | Session-start and session-end skills | #8 | Done | |
| Agent process | Agents merge their own pull requests; engine and contract changes wait for confirmation | #159 | Done | |
| Process skills, agents and hooks | Standard for skills, agents and hooks, checked in CI | #192 | Done | |
| Process skills, agents and hooks | Process agents of the project template | #193 | Done | |
| Process skills, agents and hooks | Planning skills: roadmap update, specification, ADR and sprint planning | #194 | Done | |
| Process skills, agents and hooks | Delivery skills: implement, run a sprint, pre-PR gate, commit, verify, debug, apply review | #195 | Done | |
| Process skills, agents and hooks | Review and integration skills: pull request triage, explain changes, fix CI | #196 | Done | |
| Process skills, agents and hooks | Documentation and release skills: translate, release, onboarding, retrospective, learnings, debt | #197 | Done | |
| Process skills, agents and hooks | Skills for the work this repository repeats: add a command, contract cases, new skill or agent | #198 | Done | |
| Process skills, agents and hooks | Process hooks: protect contracts, session briefing, hand-over reminder | #199 | Done |
M1 · Engine core¶
Gate: The engine replays a ledger to the same state on every platform and drives the issue cycle in a dry run on local fixtures.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| Engine foundations | Decide on ADR 0009, ADR 0010 and ADR 0024 | #46 | Done | |
| Engine foundations | Crate layout for core, runtime and adapters | #47 | Done | |
| Engine foundations | Issue step identifiers are snake_case contract values | #228 | 0001 | Done |
| Engine foundations | Merge waived by policy for small changes and staleness scoped to the issue | #233 | Done | |
| Engine foundations | Linters enforce code quality and error handling in Rust and Python | #238 | Done | |
| Engine foundations | Requirement coverage, review panel by risk tier and verification reserve in the specifications | #309 | Done | |
| State machine | Approved specification of the state machine | #11 | 0002 | Open |
| State machine | Implement the transition table and guards | #48 | 0002 | Open |
| State machine | Next step and ready work | #49 | 0002 | Open |
| State machine | Dry run of the issue cycle on local fixtures | #58 | 0002 | Open |
| Ledger | Approved specification of the ledger | #12 | 0003 | Open |
| Ledger | Append-only writer with hash chain | #50 | 0003 | Open |
| Ledger | Projection, replay and snapshots | #51 | 0003 | Open |
| Ledger | Session, issue and milestone summaries with roll-up | #52 | 0003 | Open |
| Ledger | Pull request events recorded in the ledger | #161 | 0003 | Open |
| Ledger | Work checkout with a lease and a hand-over record any agent can resume | #162 | 0003 | Open |
| Ledger | Fill the engine fields of a submitted hand-over | #274 | 0001 | Open |
| Command-line interface | Approved specification of the command-line interface and the JSON contracts | #53 | 0004 | Open |
| Command-line interface | Command schemas snapshotted with an index and a contract check in CI | #54 | 0004 | Open |
| Command-line interface | JSON envelope, error codes and exit codes | #55 | 0004 | Open |
| Command-line interface | wayloop.toml with schema and precedence | #56 | 0004 | Open |
| Command-line interface | Next, status, history and rulings commands | #57 | 0004 | Open |
| Command-line interface | Cost report per issue, phase and role | #163 | 0004 | Open |
| Command-line interface | Contract cases run through one entry point by a channel-independent runner | #221 | 0004 | Open |
| Engine documentation | Engine core concepts and command reference, English and Italian | #59 | Open |
M2 · Tracker and loop¶
Gate: Wayloop runs the issues of its own repository end to end: claim, build, verify, pull request, with gates on GitHub.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| Tracker | Approved specification of the GitHub tracker adapter | #13 | 0005 | Open |
| Tracker | Decide on ADR 0011, 0012 and 0013 | #60 | Open | |
| Tracker | Read issues, sub-issues, dependencies and milestones into the cache | #64 | 0005 | Open |
| Tracker | Write with structured block and optimistic concurrency | #65 | 0005 | Open |
| Tracker | Replace a plan item file by its tracker item | #272 | 0001 | Open |
| Loop runtime | Specification: loop runtime, worktrees and budgets | #61 | 0006 | Open |
| Loop runtime | Queue, worker pool and single ledger writer | #66 | 0006 | Open |
| Loop runtime | Worktree per issue and lane checks | #67 | 0006 | Open |
| Loop runtime | Token budgets per phase and per sprint | #68 | 0006 | Open |
| Loop runtime | wayloop run in session and unattended modes | #76 | 0006 | Open |
| Loop runtime | Failed attempts discarded with a note the next attempt starts from | #178 | 0006 | Open |
| Loop runtime | Parallel sub-agents in the session from a batch of ready work | #243 | 0006 | Open |
| Loop runtime | The loop continues at every stop until the engine reports a stop reason | #244 | 0006 | Open |
| Executors | Specification: agent executors and minimal briefs | #63 | 0008 | Open |
| Executors | Executor interface, fake executor and brief builder | #69 | 0008 | Open |
| Executors | Claude Code executor (headless and agent SDK) | #70 | 0008 | Open |
| Executors | Codex executor and generic command executor | #71 | 0008 | Open |
| Executors | Output reducers for test output, logs and structured data | #164 | 0008 | Open |
| Executors | Model and effort routing by step and complexity | #165 | 0008 | Open |
| Executors | Briefs with a stable prefix, a simplicity rule and controlled English | #166 | 0008 | Open |
| Executors | Review by a different agent, of another vendor when one is available | #173 | 0008 | Open |
| Executors | Usage limits handled by executors, in-session mode first for subscription users | #176 | 0008 | Open |
| Executors | Provider batch interfaces for non-interactive steps | #245 | 0008 | Open |
| Gates and reconciliation | Specification: gates, questions and reconciliation | #62 | 0007 | Open |
| Gates and reconciliation | Questions on issues and resumption after answers | #72 | 0007 | Open |
| Gates and reconciliation | Reconciliation of events into roadmap proposals | #73 | 0007 | Open |
| Gates and reconciliation | Hand-over comment rendered from the session summary | #74 | 0007 | Open |
| Gates and reconciliation | Reusable workflow that restarts reconciliation on merge | #75 | 0007 | Open |
| Gates and reconciliation | Deferred shortcuts become roadmap proposals | #167 | 0007 | Open |
| Gates and reconciliation | Learnings promoted to codebase patterns and agent instruction files | #172 | 0007 | Open |
| Gates and reconciliation | Notifications when the loop stops on a question or ends | #175 | 0007 | Open |
| Gates and reconciliation | Record approvals of briefs and council reports | #273 | 0001 | Open |
| Dogfooding | Run the Wayloop repository with Wayloop | #77 | Open | |
| Process policy | Approved specification of the process policy | #179 | 0021 | Open |
| Process policy | Policy registry for labels, scopes, areas, checks and rules | #180 | 0021 | Open |
| Process policy | Gated checks and label waivers as policy rules | #181 | 0021 | Open |
| Process policy | Roadmap consistency and tracker links checked by the engine | #182 | 0021 | Open |
| Process policy | Sprint lanes and required checks checked by the engine | #183 | 0021 | Open |
| Process policy | Policy results and waivers recorded in the ledger | #184 | 0021 | Open |
| Process policy | wayloop check runs the whole policy and replaces the process scripts | #185 | 0021 | Open |
| Process policy | wayloop init creates the full project template from the policy | #275 | 0021 | Open |
M3 · Living specifications and survey¶
Gate: A change on an existing codebase is surveyed, specified as a delta, implemented and merged into the living specification.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| Living specifications | Decide on ADR 0014 and ADR 0015 | #78 | Open | |
| Living specifications | Specification: living specifications and change deltas | #79 | 0009 | Open |
| Living specifications | Capability specifications, deltas, merge and archive | #82 | 0009 | Open |
| Living specifications | Consistency analysis, clarification and requirement checklists | #83 | 0009 | Open |
| Living specifications | Convergence of code and specification | #84 | 0009 | Open |
| Living specifications | Verification technique per task (TDD, BDD, end-to-end, visual, property, contract) | #85 | 0009 | Open |
| Living specifications | Judge verification for criteria that tests cannot check | #174 | 0009 | Open |
| Survey | Specification: survey scan, map and index | #80 | 0010 | Open |
| Survey | Specification: documentation with evidence and gap report | #81 | 0011 | Open |
| Survey | Scan of modules, languages, entry points and token counts | #86 | 0010 | Open |
| Survey | Tree-sitter symbols, reference graph and ranked map within budget | #87 | 0010 | Open |
| Survey | SQLite full-text and symbol index derived from files | #88 | 0010 | Open |
| Survey | Documentation with citations and unverified section | #89 | 0011 | Open |
| Survey | Surveyed specifications and gap report | #90 | 0011 | Open |
| Survey | Brownfield run on an external codebase | #91 | Open |
M4 · Upstream planning¶
Gate: A new project goes from a brief to a council report, a roadmap, user stories and generated issues, with every gate approved by a person.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| Council | Decide on ADR 0016, ADR 0017 and ADR 0025 | #92 | Open | |
| Council | Specification: expert council | #93 | 0012 | Open |
| Council | Brief from light requirements or a scope of work | #95 | 0001 | Open |
| Council | Council roles and domain configuration | #96 | 0012 | Open |
| Council | API executor for steps that do not edit code | #97 | 0008 | Open |
| Council | Council rounds and moderator report | #98 | 0012 | Open |
| Council | Proposed specialist roles with behaviour tests and usage registry | #103 | Open | |
| Decomposition | Specification: decomposition, complexity and issue generation | #94 | 0013 | Open |
| Decomposition | Roadmap proposal from the approved council report | #99 | 0013 | Open |
| Decomposition | User stories with acceptance criteria | #100 | 0013 | Open |
| Decomposition | Complexity scoring and expansion of large tasks | #101 | 0013 | Open |
| Decomposition | Idempotent issue generation with two-way status | #102 | 0013 | Open |
| Decomposition | Greenfield run from a brief to issues | #104 | Open | |
| Decomposition | Move an epic with its open stories and tasks | #271 | 0001 | Open |
M5 · Adapters and distribution¶
Gate: Wayloop installs from crates.io, PyPI and npm, and as a plugin for Claude Code and Codex, in an external repository in less than five minutes on Linux, macOS and Windows.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| SDKs | Approved specifications of the Python and TypeScript SDKs | #106 | 0014 | Open |
| SDKs | SDK with PyO3, typed API and wheels | #111 | 0014 | Open |
| SDKs | SDK with napi-rs and npx wayloop | #112 | 0015 | Open |
| MCP | Approved specification of the MCP server | #107 | 0016 | Open |
| MCP | Tool definitions generated from the command schemas | #110 | 0016 | Open |
| MCP | Server with a small tool set, budgets and read-only mode | #113 | 0016 | Open |
| Plugins and skills | Approved specification of plugins and skills | #108 | 0017 | Open |
| Plugins and skills | Claude Code plugin with skills, roles, hooks and MCP | #114 | 0017 | Open |
| Plugins and skills | Codex skills and package | #115 | 0017 | Open |
| Plugins and skills | Instructions for other coding agents generated from one source | #168 | 0017 | Open |
| Plugins and skills | Approved specification of the skill and agent catalog of the system | #200 | 0022 | Open |
| Plugins and skills | System skills for planning, from brief to issues | #201 | 0022 | Open |
| Plugins and skills | System skills for delivery, from checkout to hand-over | #202 | 0022 | Open |
| Plugins and skills | System skills for existing code and support: survey, document, gaps, debug, learnings | #203 | 0022 | Open |
| Plugins and skills | System agents shipped with the plugins | #204 | 0022 | Open |
| Plugins and skills | Agents composed per task and new project roles behind a gate | #246 | 0022 | Open |
| Distribution | Decide on ADR 0018, 0019 and 0020 | #105 | Open | |
| Distribution | Approved specification of the release pipeline | #109 | 0018 | Open |
| Distribution | Reusable GitHub Action for validation and remote execution | #116 | Open | |
| Distribution | Pipeline to crates.io, PyPI, npm and plugin marketplaces | #118 | 0018 | Open |
| Distribution | 0.5.0 preview on every channel | #119 | Open | |
| Distribution | Channel reconciler checks every channel against the envelope of the engine | #222 | 0004 | Open |
| Distribution | Decide on ADR 0028 | #231 | Done | |
| Installation guides | Installation guides for every channel, English and Italian | #117 | Open |
M6 · Benchmark and hardening¶
Gate: The benchmark results are published, the threat model is tested, performance budgets hold and no blocking defect is open.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| Benchmark | Decide on ADR 0021 and ADR 0022 | #120 | Open | |
| Benchmark | Approved specifications of the benchmark and of security and performance | #121 | 0019 | Open |
| Benchmark | Corpus of greenfield, brownfield, back-end and front-end tasks | #122 | 0019 | Open |
| Benchmark | Harness with baseline, repetitions and per-phase cost | #123 | 0019 | Open |
| Benchmark | Published results page, English and Italian | #124 | Open | |
| Security | Injection test suite and content-as-data enforcement | #125 | 0020 | Open |
| Security | Least-privilege executors and workflows, secret handling | #126 | 0020 | Open |
| Security | Sandboxed execution in a container with a cost cap | #177 | 0020 | Open |
| Hardening | Performance budgets for scan, index, next and sync | #127 | 0020 | Open |
| Hardening | Complete reference, guides and concepts in English and Italian | #128 | Open | |
| Hardening | GitLab adapter (free tier with structured fallback) | #129 | 0005 | Open |
| Hardening | 0.9.0 release candidate and contract freeze review | #130 | Open |
M7 · Version 1.0¶
Gate: Wayloop 1.0.0 is published on every channel with frozen contracts and a stability policy.
| Epic | Task | Issue | Spec | Status |
|---|---|---|---|---|
| Version 1.0 | Decide on ADR 0023 | #131 | Open | |
| Version 1.0 | Stability and deprecation policy published | #132 | Open | |
| Version 1.0 | Results on the 1.0 candidate | #133 | Open | |
| Version 1.0 | Wayloop 1.0.0 on crates.io, PyPI, npm and plugin marketplaces | #134 | Open |