A workflow operating system for AI-assisted engineering. Task state, decisions, and plans live on disk as files, not in chat history, so context survives across sessions, tools, and restarts.
This repository also configures its own agents. See what fhorja.dev tells them →
Command Claude Code
Research and recommend a technology stack for the active project by consulting official documentation, quality articles, and AAA company practices for latest stable versions. Accepts optional user-provided reference links. Produces a STACKRECOMMENDATION.md inside the active task folder with versioned recommendations…
Command Claude Code
Detect drift between TASKSTATE.md, other task-memory artifacts, and observable reality (code, tests, diff when provided), then propose the minimum set of updates so operational memory is trustworthy again. Use when TASKSTATE.md is stale or internally inconsistent after many edits, IMPLEMENTATIONPLAN.md or SLICES/.md…
Command Claude Code
Update TASKSTATE.md so it reflects the latest operational truth of the task and can be resumed safely in this or another session. Lighter-weight than state-reconcile (no cross-artifact drift detection); preferred when the update is incremental and trusted. Use when a meaningful planning, decision, implementation, or…
Command Claude Code
Ask the minimum set of high-value factual questions needed to proceed safely, then persist the result in the task repository. Distinct from decision-interview (factual gaps, not decision-driven gaps). Use when impact analysis or boundary definition revealed missing facts, correctness depends on information that is not…
Command Claude Code
Perform the terminal task lifecycle transition for a finished task: verify the spec done-conditions, set TASKSTATE.md to its final closed state, and move the task folder from active/ to archive/. The symmetric counterpart to task-init and the only official way to close a whole task. Use when every slice is closed, the…
Command Claude Code
Orchestrator-workers variant of task-init for decomposing a complex brief into N parallel independent sub-tasks. Orchestrator validates the decomposition (no overlap, clear scope per sub-task, cross-links justified); N Sonnet workers each run a full task-init in parallel producing one task folder with 5 mandatory…
Command Claude Code
Initialize the official task folder and base task memory inside projects/ /active/YYYY-MM-DD /. Creates README.md, TASKSTATE.md, SOURCEOFTRUTH.md, DECISIONS.md, IMPLEMENTATIONPLAN.md; seeds from PROJECTCHARTER.md when present; emits a multi-repo.
Command Claude Code
Provision, report, or attach a dedicated git worktree and branch for the active task on a git-backed project, so multiple tasks run in parallel on one repository without colliding on a single working tree. Opt-in and git-gated: a non-git project or an un-opted task is a no-op. Records the worktree path and task branch…
Command Claude Code
Write a simple, natural English status update for any team channel (Slack, Discord, Teams, email, GitHub PR comment, standup notes), short, grounded, and professional. Channel-portable. Use when the user wants to communicate progress quickly to teammates or reviewers, implementation or planning has reached a…
Command Claude Code
Define the smallest set of meaningful tests that protects behavior and reduces regression risk for the active task, then persist as TESTSTRATEGY.md plus a TASKSTATE.md update. Avoids forcing a strategy artifact when it adds no material signal. Detects a Godot or Unity target and routes to the matching headless runner…
Command Claude Code
Plan the Unity GameObject hierarchy and component architecture for a 3D feature before any C# is written: the scene and prefab structure, what each MonoBehaviour owns, the input model, and, for a multiplayer feature, the networked-authority declarations. Produces UNITYSCENEPLAN.md, a design-time plan an MCP-driven…
Command Claude Code
Orchestrator-workers generalization of verify-against-rubric to N=many artifacts against ONE locked rubric. Dispatches N stateless Sonnet sub-agents in parallel; each receives ONE artifact + the SAME rubric + read-only tools (no TASKSTATE, no DECISIONS, no prior history); returns structured per-criterion verdict.…
Command Claude Code
Spawn a stateless sub-agent (Claude Code Task tool, Cursor agent mode, or equivalent) with ONLY the artifact path plus the locked rubric plus read-only tools. Sub-agent returns a structured verdict (satisfied, needsrevision, or failed) plus per-criterion feedback. Distinct from self-critique-and-revise (same-context…
Command Claude Code
Verify a built web or static frontend at runtime: serve the build on an ephemeral free port, assert page identity FIRST (automatic recovery on a collision or stale server), run the web battery (overflow 320 to 2560, keyboard and focus, console errors, Lighthouse and axe when available), and decide a PASS/FAIL/BLOCKED…
Command Claude Code
Determine the current stage of the active task and recommend the single best immediate next command, editor mode, and work complexity. Routes based on TASKSTATE.md, DECISIONS.md, and IMPLEMENTATIONPLAN.md without reopening broad discovery. Use when the user wants a fast operational answer for what to do next, when the…
Command Claude Code
Macro checkpoint that assesses the real current task state against the approved plan and task artifacts, then determines what is done, what is missing, and what remains to finish the proposed work. Broader than slice closure. Returns no-op when the checkpoint would not materially change operational truth. Use when the…
Command Claude Code
Pedagogical onboarding helper that explains which command and editor mode should be used now, why, and the next 2-3 steps in a practical way for users still learning the workflow phases. Heavier than what-next; the heavier sequence is justified only when the user wants to understand the workflow, not just the next…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: