- Reworked the Junior profile from "hand-type every line" to predict-before-reveal: on each implementation task the junior commits a prediction (approach, data structure, control flow, edge cases) before the AI writes the code, then the AI grades the prediction against a rubric and names the specific gap — ownership shifts from typing the code to being able to evaluate it
- The predict-before-reveal loop runs inline as a new
/own:featurePhase 6 (Junior profile only) so the gate can't be skipped; only judgment-carrying Implementation tasks gate, while Setup and Verification tasks flow - Added a per-dimension Prediction Scorecard to the global learning registry that logs MATCH/PARTIAL/MISS over time, turning "is the junior's judgment improving?" into a measurable curve
- Added adaptive fading: the prediction gate relaxes per-dimension as the junior proves the skill (4 consecutive matches retires a dimension to a spot-check), with a deterministic staleness re-check so a faded-but-rusty dimension re-engages on its own — the gym never decays into a rubber-stamp tax
- The Junior
OWNstep now requires a self-explanation (why the actual approach is better and where the prediction broke), feeding/own:doneGate 1 with real ownership evidence - Added
docs/research/junior-profile-predict-before-reveal-validation.md— the redesign validated against learning-science literature and real-world industry practice, with prioritized refinements /own:guideis no longer the primary implementation path for juniors; it remains available for ad-hoc help (README and command docs updated to match)
- Added a boot sequence — a terminal window types your live project status (phases, task counts) on load; any key skips it,
prefers-reduced-motionbypasses it,#nobootdisables it - Added a command palette (
⌘K/Ctrl+Kwith platform-aware label, or/) for jumping between views; digits1–9jump directly and[]cycle - Added a tmux-style status bar with live segments: version, project, active phase, task totals, DoD percent, and a ticking clock
- Added a mission track to the Overview — the roadmap as a timeline with complete, active (pulsing), and roadmap-only nodes; clicking a node navigates to its phase
- Added per-phase task-burn progress bars to the Overview, derived from the same task data the kanban renders
- Added micro-interactions: pointer-tracking tile glow, count-up stats, kanban column progress bars, a segmented Definition-of-Done bar, marching-ants diagram connectors, and aurora/radar-sweep atmosphere
- Fixed the architecture diagram stretching to fill wide tiles and text overflowing its boxes — layout is now content-aware (label gutter and node boxes size to their text via JetBrains Mono's fixed metrics), with clipped text preserved as a hover tooltip
- Fixed Windows PowerShell 5.1 failing to parse the installers when run from disk — all
.ps1scripts are now pure ASCII (the same convention Chocolatey and Scoop installers use) - Expanded the render test from 35 to 48 assertions and added drift guards: every shipped command must be documented in
CLAUDE.md.template, and every.ps1must stay pure ASCII - The PR review gate now runs one full review when a PR opens and cheap incremental reviews (previous comments + newly pushed commits only) on each push, with turn caps and a one-comment praise cap
- Existing projects upgrade by replacing
dashboard.htmlonly — thewindow.PROJECTcontract is unchanged anddashboard-data.jsis untouched
- Spec-Driven Development now lives in a single browser dashboard instead of a folder of Markdown files:
dashboard.html(stable view shell, never hand-edited for content) plusdashboard-data.js(window.PROJECT, the single source of truth every/own:*command reads and writes) - The dashboard is a double-click
file://artifact — data loads via<script src>, no local server needed - Added
DASHBOARD_CONTRACT.md, shipped into every project: the schema authority forwindow.PROJECT, per-command mutation rules, the phase lifecycle, and safety invariants - Made the whole workflow dashboard-native:
/own:initfills the data file,/own:featurespecs the next roadmap phase,/own:donecompletes tasks by unique id and propagates Definition-of-Done items,/own:statusand/own:guideread it - Added the "Terminal-Futurism" default theme — a 1:1 capture of ownyourcode.dev's design system: dark near-black surfaces, terminal-green accent with glow, Outfit + JetBrains Mono, and the signature ghost-numeral motif
/own:themeregenerates only the dashboard's inline<style>and font links from a design brief — backup-first, with a post-write integrity check that auto-restores if the render contract is damaged- Installers now seed the dashboard files and retire the old
product/placeholders, warning and skipping instead of seeding a broken project when templates are missing - Added a dev-only regression suite (
bash tests/run.sh): install, headless render, and static command-contract layers - Every PR to
mainis now auto-reviewed by Claude against the OwnYourCode standards (.github/CLAUDE_REVIEWER.md) with an explicit verdict contract for branch protection - Mentioning
@claudein a PR or issue comment now summons in-thread Q&A with full PR context
- Added 4 developer profiles (Junior, Career Switcher, Interview Prep, Experienced) — the gates and quality standards stay the same, how we teach adapts to who you are
- Added profile selection to
/own:init, with settings for career focus, design involvement, analogies, and coding background - Juniors now get mandatory design involvement and momentum-driven Socratic questioning grounded in MCP research
- Experienced developers get efficiency mode, workflow preferences, and an ownership gate that asks "right approach?" instead of "can you explain?"
- The manifest gained a
profilesection;CLAUDE.md.template,/own:done,/own:feature, and/own:statusare now profile-aware
- Fixed
/own:initusing outdated internal knowledge for package versions — versions are now fetched in real time via MCP - Fixed
/own:initcreating duplicateownyourcode/directories when run from a project subdirectory
- Fixed CLAUDE.md duplication when re-running the installer
- Fixed uninstall leaving OwnYourCode files behind
- Stack recommendations now show verified versions with source attribution: read from
package.json, verified via MCP, or flagged "verify at URL"
- Added
.claude/agents/to.gitignorefor personal agent files
- Fixed the interactive prompt not appearing when installing via
curl ... | bash— input now reads from/dev/tty
- Redesigned the README as a focused landing page: "The 4 Protocols" surfaced as visible methodology, universal ownership positioning, new tagline "AI guides, you build. You own the result."
- Unified all commands under the
/own:prefix - Improved skill descriptions so skills auto-trigger more reliably
- Fixed Windows installation: PowerShell string escaping, here-string compatibility, and
irm-based project install
- Learnings now persist globally across all projects at
~/ownyourcode/learning/— patterns and failures compound across your entire engineering journey /initnow detects and teaches package managers (npm, pnpm, bun, yarn)- Package versions are always verified via Context7 + OctoCode before being recommended
- Skills now apply silently during
/featureand/done— the junior receives quality guidance without seeing skill names - Completed specs auto-archive to
completed/and task tracking updates in real time
- Added the 6 Mentorship Gates — mandatory quality checkpoints before completing any task: Ownership (can block), Security, Error Handling, Performance, Fundamentals, Testing
- Added the Learning Flywheel:
/advisefor pre-work intelligence,/retrospectivefor capturing learnings, and a persistent learning registry with patterns and failures - Added 4 fundamental skills (Testing, SEO, Accessibility, Documentation) and 2 commands (
/own:test,/own:docs) - Added the Resistance Protocol — enforced pushback when juniors try to shortcut ("just write the code for me" gets redirected to learning)
- Every completed task now produces STAR interview stories and resume bullets
- Initial release: 8 slash commands, 7 fundamental skills, 5 mentorship gates, Anti-Brain-Rot rules, Protocol D debugging, and STAR story extraction