Claude Skills by vdemkiv

Review a change.

The simple user-facing entry point for taskplane. Use whenever the user says taskplane or asks it to set up taskplane, get started with taskplane, design, build, implement, review, validate, plan, resume, or show status. The user supplies a goal and decisions; internally taskplane keeps the full strict harness: requirements and contracts, dependency-graph DoR/DoD, scoped execution, independent evidence, 26 review lenses, orchestrator-only gates, and human approval. Routes to tp-design, tp-go,...

The new-feature flow of taskplane — use when the goal is to BUILD something new: 'build a new feature', 'add X to the app', 'prototype this idea', 'build it as A/B variants', 'explore two approaches', 'greenfield this'. Enters from the product side (an idea to spec), the engineering side (a tech design to realize), or both. Front-loads a summoned north-star (strategy) check and specification refinement, renders the feature visually BEFORE and AFTER building, and can build the same requirement...

The pre-build solution-design flow of taskplane. Use when the user says 'taskplane design', asks to design a new feature or approach before coding, wants a technical design based on requirements and/or an existing codebase, wants alternatives and trade-offs, or needs dependency/contract/rollout decisions settled before Build. Produces an approved Design Contract with proposed graph, Design DoR/DoD, solution-design evidence, and conditional visualization. It designs the HOW; it does not implem...

The engineering persona of taskplane — owns whether the built thing is right and sound. Use for validating completed work: 'review this' (PR/branch/code/work), 'security review', 'architecture review', 'does this match the requirement', 'what depends on X', 'blast radius', 'run the retro', 'sign-off'. Engineering consumes Evaluate's sealed direct evidence, launches no lens workers, and returns missing or insufficient substantive evidence to a fresh zero-lens Evaluate judgment. It also walks r...

The internal delivery driver behind the taskplane facade — goal-shaped asks ('build X', 'fix X', the word taskplane) land on the facade, which routes here. Reach for this skill directly only when the user explicitly drives the loop: 'start governed work', 'run the loop', 'run tasks in parallel', 'dispatch the wave', 'run the retro', 'log tech debt'. Drives governed delivery end to end, routing to the right persona — tp-product (define WHAT), tp-design (propose HOW), tp-build (realize), tp-eng...

Use when the user asks how taskplane works or how to get started: 'taskplane help', 'how do I use taskplane', 'what can taskplane do', 'taskplane tour', 'getting started with taskplane', 'what is taskplane', 'I installed taskplane, now what', 'taskplane is not doing anything'. Gives the mental model, the quickstart, and which skill to reach for.

The on-demand STRATEGIC lens of taskplane — the north-star review. Summon it, don't schedule it: 'north-star this', 'board this', 'strategic take on this', 'does this serve the direction', 'is this worth doing given where we're going', 'strategic review of this task, diff, PR, plan or idea'. It is NOT a loop stage — tp-product (the WHAT) and tp-engineering (the SOUND) are the two automatic seats; this is the third lens you call when you want a direction check. Read-only and advisory: it measu...

The product persona of taskplane — owns the WHAT. Use for anything about what to build and whether it's the right thing: 'spec this', 'write acceptance criteria', 'refine the requirement', 'change request', 'should we build this', 'prioritize', 'log tech debt (product)', 'record the decision'. Authors and scores requirements, closes refinement gaps, and holds the plan-approval recommendation. Strategy/direction calls ('given where we're going, is this worth it') belong to the summoned north-s...

Use when the user asks where things stand with taskplane-governed work: 'status', 'where are we', 'what's the state of the loop/track/requirements/debt'. Read-only snapshot rendered as the inline mission-control dashboard: active track, loop step, tasks, open requirements, tracked debt, KB size, dependency graph — with an explicit action banner (gate buttons if a decision is yours, 'no action needed' if agents are working).

Run taskplane governed work inside Claude Tag (Slack). Use when a session is running as Claude Tag in a Slack channel or thread and the team wants governed work there - 'run this with taskplane', 'governed task', 'plan first and let the channel approve', 'taskplane in Slack'. Adapts the loop to Tag's environment - repo-persisted knowledge store for the ephemeral sandbox, human gates answered by real people replying in the thread, dashboard posted to the channel. Never self-approves a gate.