agentleFS
gleicongleicon/devskills/skills/ds-roadmap/SKILL.md

ds-roadmap

Skill20 starsChanged 46 days agoMITView on GitHub

Turn a goal or another command's output into an ordered task roadmap.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 5 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When invoked, sequence work into a roadmap checklist.

    line 7
  2. Distinctive:

    Use whatever the user provides as the source of tasks:

    line 11
  3. Distinctive:

    If nothing is provided, ask once for the goal, then proceed.

    line 16
  4. Distinctive:

    Write them as a status checklist, appending to existing tasks rather than discarding them:

    line 22
  5. Distinctive:

    Before finalizing, show the breakdown and ask whether the granularity is right — merge or split on the user's steer, don't just emit and …

    line 24
  6. Distinctive:

    Keep it terse. No estimates, no owners, no ceremony.

    line 32

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

When invoked, sequence work into a roadmap checklist. You order the work; you do not choose the architecture or make technical decisions — those are the engineer's. The companion to /ds-spec: spec defines the WHAT, this orders the work to get there.

Input

Use whatever the user provides as the source of tasks:

  • a goal or feature description,
  • a SPEC.md,
  • pasted output from another command (e.g. /ds-code-quality-review findings, a list of bugs).

If nothing is provided, ask once for the goal, then proceed.

Process

  1. Resolve the target file: .project/roadmap.md if .project/ exists, else roadmap.md in the current directory. Read it first if it already exists. Do not create .project/ — write to the current directory when it's absent.
  2. Break the input into the smallest vertically-sliced tasks — each cuts a narrow but complete path through all layers (schema → logic → API → UI → test), shippable and demoable on its own. A single-layer horizontal slice ("all the schema", "all the endpoints") is the anti-pattern — it can't be shipped or verified alone.
  3. Write them as a status checklist, appending to existing tasks rather than discarding them:
    • [ ] todo · [~] in progress · [x] done
  4. Before finalizing, show the breakdown and ask whether the granularity is right — merge or split on the user's steer, don't just emit and move on.

Rules

  • Order and scope only. No library choices, no patterns, no architecture — record what the engineer decides, don't decide it.
  • Tasks are outcomes ("parser rejects malformed input"), not instructions on how to implement them.
  • Sequence enabling cleanup first: if a prefactor makes the real change easy, make it its own early task — make the change easy, then make the easy change.
  • Wide mechanical refactors (rename a shared symbol/column that breaks every call site at once) don't fit one vertical slice. Sequence expand → migrate → contract: add the new form beside the old, migrate call sites in blast-radius-sized batches (each shippable, CI green), then delete the old last, only after every migration batch has landed.
  • Keep it terse. No estimates, no owners, no ceremony.
  • The roadmap is the whole file and you are its only writer — nothing else lives there, so there is nothing in it to preserve or work around.

Output

Display the updated roadmap and the path.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.