agentleFS
AJschoonsAJschoons/robowright/skills/work-planner/SKILL.md

work-planner

Skill1 starsChanged 9 days agoMITView on GitHub

The front door for a single ask — turns one task, idea, or dictated request into a work order a fresh-context agent can run with. Use when the Owner says "plan this", "scope this", or "write this up for an agent", or hands over an ask too loose to just do and too small for a project (`planner`).

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When two readings of the ask would send step 2 in different directions, ask that one question first.

    line 24
  2. Distinctive:

    Ask for the two reports in GROUNDING.md, about this ask.

    line 29
  3. Distinctive:

    Read LEADING-WORDS.md: use its word when one fits, and add a row when the outside report names a better one.

    line 43

≈ “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

Work planner

One ask in, one work order out: a document that survives a cold start. An agent that has never seen this conversation reads it and runs.

1. Name the destination

State the destination back in one or two lines: what exists when the work is done, and its kind — build, fix, decide, write, investigate. The Owner may dictate; read through slips to the intent.

The destination sizes the ask:

  • Mechanical — a rename, a move, a config tweak, a fix whose cause is known. Step 2 runs the inside report only.
  • A project — more than one session can hold, or fog a single grilling will not clear. Say so, hand the Owner to planner, and end here.
  • Everything else runs every step.

When two readings of the ask would send step 2 in different directions, ask that one question first. Otherwise carry on; the destination goes out again as the first Assumed line of step 4.

2. Ground

Ask for the two reports in GROUNDING.md, about this ask.

Work step 3 while they run. Save both reports beside the work order as <slug>-grounding.md.

Done when every fact the plan leans on has a citation, or is marked as belief.

3. Translate

Invoke domain-modeling. Every noun in the ask resolves to a glossary term or the code's own symbol (SyncStore, where the Owner dictated "sink store"). A noun that stays fuzzy or overloaded becomes a question for step 4.

Then the patterns. Wherever the ask spends a sentence describing an approach, find the leading word the model already holds for it ("one thin path end to end first" is a tracer bullet). Read LEADING-WORDS.md: use its word when one fits, and add a row when the outside report names a better one. A word earns its place when it retires a sentence and changes what an agent does by default.

Done when every concept in the draft has exactly one name, and that name is a glossary term, a code symbol, or a leading word.

4. Grill

Invoke grilling on what is still open. The destination, the sizing call, and every term you settled yourself open the round as Assumed. The dissenting view goes to the Owner as a question whenever it would change the approach. The decision record lands in the work order.

Done when the frontier is empty.

5. Write the work order

Fill WORK-ORDER-TEMPLATE.md, applying writing-for-agents: completion criteria that are checkable and exhaustive, targets phrased as the positive. It goes in the project folder when the ask belongs to one, otherwise the inbox the profile → Paths names, as <YYYY-MM-DD>-<slug>.md.

Then run the cold start: hand a fresh subagent the work order alone and ask what it would do first and what it would have to ask. Each question it raises is a gap. Close the gaps in the work order, once.

Done when the cold-start reader's questions are all answered by the document.

6. Hand over

Give the Owner the path, the destination line, and anything under Still open. Three ways on:

  • Here — the work fits what is left of this session and they say go: carry it out from the work order (bdd first for a behavior change).
  • A fresh session starts from the file. The work order is the whole hand-off.
  • Unattended, under Robowright's gates — where planner is installed, a fresh Planner session turns the work order into a brief: /planner order <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.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

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.