▸ Agent Skills

Planning

Turn ideas into specs, tickets, and multi-session plans.

Sections

  • estimate-work — Size and estimate a body of work to set expectations and sequence delivery - turn a plan or ticket set into defensible estimates with ranges, named assumptions, and surfaced risk instead of false-precise point numbers. Use when the user wants to estimate or size work, forecast how long something will take, choose between relative sizing and time-based estimates, calibrate against past delivery, or decide whether to spike before committing.
  • grill-me-with-docs — Relentlessly interview the user to stress-test a problem, architecture, or design from scratch, while actively maintaining a domain glossary (CONTEXT.md) and recording major architectural decisions as ADRs. Use when starting a new initiative, tackling ambiguous domain problems, or when the user asks to grill with docs or ADRs.
  • grilling — Interview the user relentlessly, one question at a time, to stress-test a plan, design, or architecture before any code is written. Use when the user asks to 'grill me', 'poke holes in this', 'stress-test this plan', or 'interview me about this design'. Walk every branch of the design tree, look up facts in the codebase, and put each real decision to the user with a recommended answer before proceeding.
  • handoff — Compact the current conversation into a handoff document for another agent to pick up.
  • prioritize-backlog — Groom and prioritise a backlog - order, cut, cluster, and sequence a pile of work by value against cost and risk so the next thing to do is obvious. Use when the user wants to prioritise or rank a backlog, decide what to build next, trim or triage a pile of ideas, apply a lens like value-vs-effort, RICE, or cost-of-delay/WSJF, or sequence work around dependencies.
  • software-factory — Run a feature through four approval gates - Product, Architecture, Program Design, Slices - so every decision that matters is made and signed off before implementation code exists.
  • store-plan — Capture the current conversation's plan, decisions, and action items into a structured, reviewable markdown file in the project's docs/plans/ directory.
  • tars-goal — Set or resume a persistent goal and pursue it until every requirement is verified or a blocker prevents further work. Use when the user invokes /tars-goal or explicitly asks to establish or resume a persistent goal.
  • to-spec — Turn the current conversation into a spec (sometimes called a PRD) and publish it to your project's issue tracker - no interview, just synthesis of what you have already discussed.
  • to-tickets — Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, and publish them to your project's issue tracker - as one file per ticket locally, or as native blocking links on a real tracker.
  • wayfinder — Plan a huge chunk of work - more than one agent session can hold - as a shared map of investigation tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.