author-and-verify-extension
Skill2 starsChanged 27 days agoMITView on GitHub
Builds a new sushiAI extension end to end: designs the manifest, validates it with validateExtensionManifest, builds, and proves it loads in a real Electron smoke run against a throwaway extensions directory. Use when asked to add, scaffold or build an extension, plugin or surface.
- Reads credentials
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- autoreviewSkill
- cut-releaseSkill
- deslopSkill
- extension-manifest-changeSkill
- figma-to-codeSkill
- ship-prSkill
- sushiai-taskSkill
- sushiai-testingSkill
- sushiai-uiSkill
- ui-evidenceSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
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
- Very specific: line 32
Validate it: before running anything: validateExtensionManifest from electron/extensions/manifest.cjs (Node, required directly) against { …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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.
Files
Author and verify an extension
This is the concrete answer to "I want to write a plugin" — the workflow a
contributor actually runs, not just a manifest reference. See
AGENTS.md's "Extension contract is the capability surface" section for the
contract facts cited below; this skill is the procedure for exercising it.
The loop
- Design the manifest shape from what was asked: which surfaces
(with
instancePolicy,stateScope, aview.documentdescribing fields and views), whichnavigation/actionsentries expose it, whichcommandsopen it. Constraints from the contract:MAX_VIEWS = 1(onlyviews[0]renders),commands[].surfaceIdis required, labels onnavigation/actionsmust be unique within the manifest,editablefields must already exist in that view'smeta. - Write the manifest to a throwaway directory — e.g.
/tmp/sushiai-ext-<name>/manifest.json— never intotests/fixtures/extensions/. That directory is a tracked fixture set;scripts/smoke.mjsasserts its entry count matches exactly what's checked in, so dropping a scratch extension there fails an unrelated assertion. - Validate it before running anything:
validateExtensionManifestfromelectron/extensions/manifest.cjs(Node,required directly) against{ ...manifest, source: { kind: "local", path: "<the throwaway dir>" } }. Fix every thrown error before moving on — the validator's message names the exact field. - Build, then launch against the throwaway extension with a driver script
you write, not
scripts/smoke.mjsitself:npm run buildfirst (the app under test is the built bundle - a staledist/silently tests the previous build).scripts/smoke.mjscannot be redirected by settingSUSHIAI_EXTENSIONS_DIRin your shell - it unconditionally overwrites that key to the literal"tests/fixtures/extensions"when it spawns Electron (env: { ...process.env, SUSHIAI_EXTENSIONS_DIR: "tests/fixtures/extensions" }), so an external value is silently discarded. Instead, copyscripts/smoke.mjs's own setup (the build-freshness check and itselectron.launch(...)call, first ~75 lines) into a throwaway script, point your copy'sSUSHIAI_EXTENSIONS_DIRat the throwaway directory, and replace its probe-specific assertions with assertions about your own extension (its nav label appears, its surface opens without a renderer error). - Report pass/fail with what was verified: the extension loaded, its surface opened, and (if state is involved) a write round-tripped.
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
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.

