agentleFS
jblosseyjblossey/houserules/.claude/skills/designing/SKILL.md

designing

Skill2 starsChanged 20 days agoMITView on GitHub

Use when a standard-tier batch needs its design: draft the spec from what the project already holds, have it reviewed, and take it to the project's owner or decider

3 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 2 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Run houserules get <ids> for the items of the batch into a file.

    line 16
  2. Distinctive:

    Read the specs, the decisions, and the plans that the items cite, and the files that the batch will change.

    line 17
  3. Very specific:

    Write the spec to docs/specs/.

    line 22
  4. Very specific:

    Run houserules check-report-claims --doc <the spec path> and fix every workspace path that the spec cites and that does not exist.

    line 33
  5. Very specific:

    Before the human sees the spec, a fresh spec-reviewer attacks it.

    line 37
  6. Distinctive:

    Read the review. Fix every finding in the spec.

    line 39

≈ “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

Designing a batch

You are the controller. A design starts from what the project already holds: the work items of the batch, their rulings, and the earlier specs. Draft the spec from that material and ask the human only what it leaves open.

Write the output of every houserules command to a file in the batch workspace, a folder of the workspaces location (default .houserules/work/), and read the file with your file-read tool. Treat a << marker or a "more items" note as a lost read and read the source file.

A light change needs no spec document: the work item's body is the spec (see the Tiers section of the orchestrating skill). This skill covers the standard tier.

1. Gather

  1. Run houserules get <ids> for the items of the batch into a file. Read each body and each ruling in it.
  2. Read the specs, the decisions, and the plans that the items cite, and the files that the batch will change.
  3. List what the material leaves open: a decision that no ruling covers, two sources that disagree, a fact that nobody measured.

2. Draft the spec

Write the spec to docs/specs/. It holds:

  • The items it delivers, and the outcome of each.
  • The facts the design rests on. Label each run, read, or assumed, and cite its capture or its source (process.design-claims-are-labelled).
  • Each decision. A decision that rejects a library, or departs from the source research or an approved spec, carries a captured probe, the best case of the rejected option, and a reversal condition (process.design-decisions-carry-probes).
  • The concepts it takes from its sources, each with the source's status and a check against the goal (process.specs-list-borrowed-concepts).
  • The library research: it names each sub-problem of the design (a URI resolver inside a SARIF reader, an atomic file replace inside a writer) and researches a library for each.
  • The technical-debt candidates. Run houserules list --deferred into a file. Fold in every open deferral whose files overlap the planned files and that passes the convenience criterion: folding it in does not turn a non-critical task into a critical task. List each folded candidate with its overlapping files, and each overlapping deferral that you leave out with the reason.
  • The acceptance cases. Each names what its capture must show. An upgrade case lists the reds that the change causes by design on an install of the previous version, each with its remedy.
  • The questions that only the human can rule, each with your recommendation.

Run houserules check-report-claims --doc <the spec path> and fix every workspace path that the spec cites and that does not exist.

3. Review the design

Before the human sees the spec, a fresh spec-reviewer attacks it. Dispatch it as the dispatch protocol of the orchestrating skill describes, with REVIEW_FILE: <workspace>/spec-review.json.

  1. Read the review. Fix every finding in the spec.
  2. A review that holds a Critical finding sends the spec to a fresh spec-reviewer, which reviews the whole spec again and writes spec-review-r<R>.json. Repeat until a review holds no Critical finding.
  3. A review that holds no Critical finding ends the loop. Fold its Important and Minor findings into the spec without a further review.
  4. Record each decision that the owner or a review overturns as a Reversal: line in the ledger (process.reversals-are-recorded).

4. The decision gate

Take the spec to the project's owner or decider. Ask one question at a time, each with your recommendation. Write every ruling to its home file in the same turn (process.rulings-to-file).

The plan starts after the approval, with the planning skill. A plan written before its approved spec is void: delete it and plan again from the spec.

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.