agentleFS
DoistDoist/todoist-ai/.agents/skills/build-feature/SKILL.md

build-feature

Skill565 starsChanged 5 days agoMITView on GitHub

Builds a new tool, parameter or behaviour in thin test-first slices, after settling the open decisions with the owner. Use when asked to "add", "build", "implement", "support" or "create" something, such as a new tool or a new tool parameter.

17 other projects use a version of this, 2 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 16 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

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

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

    Write each behaviour as an input and the expected output, for example: "update-tasks with deadlineDate: "remove" sends deadlineDate: null …

    line 18
  2. Very specific:

    Follow verify. If the feature needs restructuring first, do that as a separate change with refactor.

    line 41

“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “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

Build a feature

1. Read the closest existing feature

Find the tool most like the one you are building and read it with its test: src/tools/add-projects.ts for a batch write, src/tools/find-tasks.ts for a read, src/tools/update-tasks.ts for an update. Read the rules in AGENTS.md ("Tool Schema Design Rules", "Adding a New Tool"). Done when you can name the file you will copy patterns from.

2. Settle decisions in one round

List what the request leaves open: naming, input shape, what to return, error cases, annotations (readOnlyHint, destructiveHint, openWorldHint), whether it needs a routing line in src/mcp-server.ts. Ask the owner all of them at once, each with your recommendation. Don't ask what the code can answer. Done when every decision has an answer.

3. Write "done" as checkable behaviours

Write each behaviour as an input and the expected output, for example: "update-tasks with deadlineDate: "remove" sends deadlineDate: null to updateTask". Each becomes a test. Done when a test could pass or fail on every one.

4. Build in thin slices, test first

For each behaviour: write its test at the public interface (<tool>.execute(args, mockTodoistApi)), see it fail, make it pass, run its test file. Follow write-tests. Done when every behaviour from step 3 has a passing test you saw fail.

5. Update everything the change touches

You changedAlso update
Added a toolsrc/utils/tool-names.ts, src/tool-registry.ts, src/index.ts (named exports), src/tools/tool-annotations.test.ts
A tool writes a list of itemsPer-item failures (see "Batch write tools" in AGENTS.md); batch-failures-contract.test.ts checks it
A clearable optional fieldThe "remove"/"unassign" string, both runtime branches, the schema description
Tool or field descriptionsnpm run eval -- --label before, then --label after; diff tmp/eval/*.json (paid models: ask first)
The tool surface grewsrc/token-footprint.test.ts may fail: raise DEFAULT_TOKEN_BUDGET only if needed, and say so in the PR
Cross-tool routingThe instructions string in src/mcp-server.ts, only if a model needs it to pick between tools
Output schemassrc/utils/output-schemas.ts; never .nullable() (output-schema-nullability.test.ts)
Structure (new folder or helper)CODEBASE.md, by its "Keeping CODEBASE.md accurate" rules

Done when each row that applies is handled.

6. Verify and report

Follow verify. If the feature needs restructuring first, do that as a separate change with refactor.

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.