tailrocks/tailrocks-roadmap-skills
Evidence-bound roadmap delivery from idea capture and research through shaping, planning, execution proof, reconciliation, and retrospection.
Shapes a DRAFT or SHAPING roadmap item through a one-question-at-a-time interview, and writes every answer into the item as it resolves. Use only when the user explicitly requests this skill with a live human. Do not use for final readiness (tailrocks-finalize), for a READY item, or without a live human.
Closes the shaping interview on a SHAPING roadmap item: resolves every screen, flow, and open question, then grants READY. Use only when the user explicitly requests this skill with a live human. The only source of READY. Do not use on a raw DRAFT (tailrocks-brainstorm first) or without a live human.
Stress-tests an idea, plan, or decision through a challenge conversation before action. Use when the user asks to be grilled, challenged, interrogated, or stress-tested. Conversation-only: the skill retrieves facts, recommends answers, leaves decisions to the user, requires confirmation, and never executes or writes artifacts.
Captures one raw product or feature idea as a new DRAFT roadmap item under roadmap/<slug>/ and registers it in the index. Use only when the user explicitly requests this skill with raw words: a sentence, a paragraph, or pasted notes. Capture only: no interview, no research, no plan. Do not use for an existing item or for a verified finding.
Converts a READY roadmap item into roadmap/<slug>/plan/ and goal/: a coverage ledger, a research-gap manifest, an OpenSpec-grammar spec, one zero-context plan per work item, and the goal handoff. Use only when the user explicitly requests this skill with a READY item. Do not use on unshaped items or on one-session changes.
Executes every surface that a roadmap item claims to ship, confirms or refutes each reported defect, and writes the verification round with subagent fan-out, evidence per surface, and a vacuous-proof audit. Use only when the user explicitly requests this skill. Judges only: never fixes, never writes status.
Trues up roadmap/<slug>/ with execution reality: re-runs each plan row done criteria, rejects criteria that executed nothing, folds the newest verification round into the Remaining of the item, and sets the supported status. Use only when the user explicitly requests this skill. Only this skill sets DONE.
Records one user decision on a roadmap item: validates it against settled ground, dates it with its reason, propagates it, and flags what it invalidates, including the reopen of READY or PLANNED work. Use only when the user explicitly requests this skill with one stated decision. Never decides for the user.
Captures what a user found wrong with the shipped work of a roadmap item: their words verbatim, one statement per defect, reproduction as given. Use only when the user explicitly requests this skill. Captures only: never investigates, judges, or fixes. tailrocks-prove verifies the round.
Runs deep, sourced research into a reusable topic under research/, for a question or to extend a roadmap item, with parallel investigators. Use only when the user explicitly requests this skill. Do not use for decisions only the user makes, or for questions that one lookup answers.
Rebuilds which skills ran on a shipped roadmap item from commit trailers, diffs that history against its Decisions, Must not, spec IDs, and verification rounds, and proposes patches to the skills at fault. Use only when the user explicitly requests this skill. Proposes only: tailrocks-skill-update edits skills.
Converts one already-verified finding or one approved standalone plan into one DRAFT roadmap item on its delivery branch and pull request. Use only when the user explicitly requests this skill with verified input. Writes roadmap artifacts only. Never implements, never seeds raw ideas, and never touches an existing item.