Milestones
List view
Harness - a mutant counts as killed only when a test failed (broken import, slow run) - tool-profile floors raised to the scores reached - a branch mutates only the lines it changed - every CLI test runs in a throwaway HOME - real-binary scenarios: plugin already installed, two projects on one HOME - Biome checks a file the moment an agent writes it - a gate a tree already passed: skipped locally, reused on promotion Fixes - clean removes marketplaces aidd never registered - an empty tick restores everything - the MCP exclusion strips nothing from a real manifest - telemetry off leaves a foreign hook non-executable - a worktree hook writes lefthook's example config into cli/ - an empty worktree id reads as stated; a record dated from its text - unquoted plugin root kills every hook on Windows - OpenCode loads journal.cjs from a path its build never writes Done when the harness is in place and every fix is on next. --- _Generated by [Claude Code](https://claude.ai/code)_
Due by September 17, 2026•29/29 issues closedSomeone who is not us installs the plugin, turns measurement on, works a week, asks what it cost, and gets a figure they can act on — on their own machine, their own tool, their own OS. Today the chain is proven against captured payloads on CI; nothing proves a real session on a real Windows or Linux box writes what the reader expects, and nothing proves the figure still holds after a hundred sessions. Everything else in the telemetry backlog adds an axis to a report that already works, and ships after.
Due by September 10, 2026•1/1 issues closedMake every AI session traceable from the commit it produced to the cost it carries. The framework emits what providers never expose (mode, effort, skills, task), the CLI owns installing and checking that pipeline, and no commit is ever blocked by it. Done when a commit created by an agent carries its session id on every supported tool, and a single command proves whether telemetry is on and correctly wired.
Due by August 21, 2026•34/34 issues closedFix three fronts where AIDD no longer behaves as documented: skills distributed to Codex, skill behaviour from one call to the next, and rules duplicated between skills and project memory. Done when every skill starts and behaves the same on every tool, and every rule has a single home.
Due by August 13, 2026•4/4 issues closedAllow users to create, manage, ask about an agile backlog with all types of items : epics, stories, defect, tasks, backlog, etc
Due by August 6, 2026•1/1 issues closedClarify three user-facing flows that currently leave ambiguity: PM spikes become a written artifact, brainstorm gets an explicit discovery flow, and CLI status exposes the documented machine-readable output.
No due date•5/5 issues closedMigrate the CLI repository inside Framework to help user to install Framework properly and maintain it in time
Due by July 23, 2026•3/3 issues closedFix the base a skill is built from: the rules and the template every new skill copies. Today the template still teaches the format onboard abandoned, so nothing propagates. Once the base carries the load budget, project-memory and learn are the first two skills held to it, and every later cleanup follows without a new decision.
Due by July 17, 2026•2/2 issues closedHelp user to onboard in the Framework. Skill /onboard + documentation
Due by July 10, 2026•6/6 issues closedTransformation du framework monolithique en architecture plugin composable. Chaque bundle (dev, pm, qa, etc.) devient un plugin installable indépendamment via le CLI AIDD.
No due date•8/8 issues closed