Groom Roadmap
Productivitygroom-roadmapGroom Roadmap is an Agent Skill for productivity published in the talkincode/toughradius repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Roadmap and development-plan self-iteration (grooming). After each subtask delivery, or during periodic maintenance, check off completed items, update milestone status, backfill / split / re-prioritize subtasks, and fold newly discovered needs in via the feature checklist to keep the plan self-consistent and aligned with the checklist.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/talkincode/toughradius/blob/HEAD/.agents/skills/groom-roadmap/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/groom-roadmap, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Skill: Roadmap / Plan Self-Iteration (Grooming)
Scope:
TR-F022| Usually invoked by../orchestrate-roadmap/SKILL.mdafter delivery; can also run standalone.
When to use
- After a milestone subtask is delivered (PR merged or ready).
- During periodic roadmap maintenance, when a subtask is too large / outdated / already implicitly done.
- When implementation surfaces a new gap, new dependency, or a change in requirement boundary.
Steps
- Check off delivery: change the delivered subtask
- [ ] M*.*to- [x]; sync the milestone overview table and the milestone section status (planned -> in progress -> delivered; delivery means merged tomain+ CI passing). - Backfill learnings: based on the round's PR result / blockers,
- split follow-up work discovered during implementation into new
- [ ] M*.*subtasks (keep each independently deliverable, revertible, verifiable); - split oversized subtasks; annotate removals with the reason;
- mark items with an external dependency / pending decision as
blockedwith the reason.
- split follow-up work discovered during implementation into new
- Re-prioritize: keep
M1 -> M2 -> M3by default; adjust only when dependencies / blockers / new delivery evidence change, and explain the reason in the PR. - Align scope: route newly discovered needs through
../align-feature-checklist/SKILL.md- schedule only what maps to aTR-F; reject anything hittingTR-N001-TR-N005; never silently expand the roadmap. - Consistency check: ensure the milestone overview status <-> subtask checkboxes <->
docs/feature-checklist.mdstatus are consistent; keep the CN/EN checklists in sync; no dangling references to deleted files / skills. - Commit: plan-doc changes go through a PR - either folded into the delivery PR or as a small follow-up PR; reference the milestone ID.
Boundaries
- Only edit plan / scope docs (
docs/roadmap.md,docs/feature-checklist.mdand its English version, the in-session plan); do not change product code in this skill. - Do not invent tasks: every new subtask must anchor to a
TR-Fand be a minimal closed loop. - Do not relax guardrails: non-goals, PR-only, and quality-gate requirements stay unchanged.
Acceptance
- Delivered subtasks are checked off; the milestone status table is in sync
- Added / split / blocked subtasks all anchor to a
TR-Fand are independently verifiable - New needs are aligned via the feature checklist (CN/EN in sync); no
TR-Ntouched - Roadmap <-> feature checklist status are consistent; no dangling references