agentleFS
requierequie/endostack/skills/typescript-best-practices/SKILL.md

typescript-best-practices

Skill4 starsChanged 15 days agoMITView on GitHub

Sixteen type-safety rules for writing and reviewing TypeScript, with do and don't examples in references/patterns.md. Use whenever you open or edit a .ts or .tsx file, or on /typescript-best-practices.

77 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use setup-pstack, poteto-help and 2 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • bnemaSkill

    setup-pstack

    Configure which Pi models pstack uses per role. Detects available provider/model specs and writes pstack's user configuration. Use for /setup-pstack, "configure pstack models", or changing pstack model choices.

    Also in 22 projects that use this

  • ianclearySkill

    poteto-help

    Guides users through pstack setup, /poteto-mode, and picking the skill, playbook, or principle for a task. Type /poteto-help with a question.

    Also in 12 projects that use this

  • tristanphvnSkill

    ketchup-video

    Catch the user up like /keel:ketchup, with a short silent video of slides for what happened and the waiting calls as text under it. Use for /keel:ketchup-video.

    Also in 8 projects that use this

  • instructureSkill

    ketchup

    Rebuild context after /clear by analyzing uncommitted changes across Canvas LMS and all plugin repositories

    Also in 17 projects that use this

The file

We split this skill into the rules it gives an agent

4 rules. 0 are common, 1 is distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 26 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Read principle-type-system-discipline (../principle-type-system-discipline/SKILL.md) first.

    line 9
  2. Very specific:

    If your agent supports path-scoped skills, attach this one to /*.ts and /*.tsx.

    line 11
  3. Distinctive:

    Before you write anything, find the repo's existing conventions: its schema library, its logger, the name it uses for union tags, and the …

    line 13
  4. Very specific:

    Derive types from generated sources: When types already come from protobuf, OpenAPI, GraphQL, or database migrations, build on them with …

    line 32

“Common” is measured: kept by at least 6 in 10 of the 26 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

typescript-best-practices

Read principle-type-system-discipline (../principle-type-system-discipline/SKILL.md) first. These rules apply it to TypeScript. references/patterns.md has a short code example for each rule, numbered to match.

If your agent supports path-scoped skills, attach this one to **/*.ts and **/*.tsx. Otherwise the router or the user loads it when TypeScript is in play. For other languages, use principle-type-system-discipline and principle-boundary-discipline (../principle-boundary-discipline/SKILL.md) directly.

Before you write anything, find the repo's existing conventions: its schema library, its logger, the name it uses for union tags, and the shape of its brand marker. Follow them over the examples in the reference.

Modeling

  1. Tagged unions for variants. Model alternative states as a union of object types that share one literal tag field. A boolean flag plus optional fields lets contradictory combinations exist (loading: true with an error set). Use one tag name across the codebase.
  2. Branded primitives. Intersect a primitive with a readonly brand property so an InvoiceId cannot be passed where a CustomerId is expected. Mint the brand only through a type guard that validates (value is InvoiceId) and a parse function that wraps the guard and throws on failure. Never brand with as. Check once where the value enters; downstream code trusts the brand.
  3. Make illegal values unconstructible. Prefer a shape that cannot hold a bad value to a runtime check: [T, ...T[]] for a non-empty list, Array<[K, V]> for paired data, a start plus a duration instead of start and end with an ordering invariant in a comment. Provide helpers for derived views (the computed end time). Brand the duration only if some unrelated number could be passed in its place.
  4. Use the simplest type that keeps every operation total. A plain array is fine while every operation on it works for empty input (a sum of nothing is zero). Strengthen to non-empty only where the weak type forces a lie: a !, an index cast, or an "unreachable" throw. Or widen the return type to include undefined. Provide a guard so a plain array can be checked once and then carried as non-empty.

Untrusted data

  1. External data is unknown, never any. RPC payloads, parsed JSON, postMessage events, IPC, files, environment variables, and database rows all arrive as unknown and get narrowed before use. any turns off the checker exactly where an attacker controls the input.
  2. Schemas before hand-written guards. Check whether the codebase uses a runtime validation library and already defines a schema for this data. Define one schema and infer the static type from it. Use the non-throwing parse call when bad input is an expected outcome (a user form), and the throwing one when it is a bug. Never add a validation library for one guard. If the static type already exists, annotate the schema with it so the compiler rejects a schema that checks less than the type promises.
  3. No as except after validation. When you remove an assertion, find out why inference failed: add a missing discriminant, narrow a source typed as Record<string, unknown>, parse at the untyped boundary with the owning schema, or, if the fact truly cannot be expressed, use a brand or satisfies.
  4. Narrow in this order, best first: switch on the tag; an in check; typeof or instanceof; a custom guard; an assertion after validation.
  5. Guards must check what they claim. A guard that lies is worse than as because its name promises safety. Name guards isX or hasX. Prefer tag narrowing when a tag exists.
  6. Exhaustive switches. In the default branch of a switch over a union, assign the value to a local typed never. Adding a variant then fails to compile at every switch that forgot it. Return that local in value-returning switches; mark it used with void in statement switches.
  7. satisfies over as for literals. It checks conformance and keeps the literal types.
  8. Parse once at the boundary. Data crossing into the system is parsed once into a named domain type. A loose record type never travels past that parse. Wire formats tolerate unknown fields so older code can read newer messages. Persisted JSON is a versioned blob parsed inside try/catch. Inner functions do not re-validate. See principle-boundary-discipline.
  9. Derive types from generated sources. When types already come from protobuf, OpenAPI, GraphQL, or database migrations, build on them with Pick, Omit, Partial, and typeof. A hand-written duplicate interface drifts the first time the source changes.

Calls, tests, and telemetry

  1. One object parameter over several positional ones. Swapped positional arguments of the same type compile without complaint; named fields do not. Exempt hot paths (per-frame rendering, tokenizers, parsers, tight loops) where the allocation shows up in a profile.
  2. Run real code in tests. Do not mock what can run on a dev machine. Use the test framework's own facilities, including its checks for leaked handles and undisposed resources. Check UI behavior in a real running build. Keep mocks for dependencies that cannot run locally. See the tdd skill (../tdd/SKILL.md).
  3. Structured telemetry. Log through the project's structured logger with enough fields that an id alone lets someone debug the case. No bare console.log in shipped code. Never log secrets, tokens, or full request bodies that may contain them.

Done when

No any touches external data, every as follows a validation, no guard lies, every switch over a union is exhaustive, and boundary data is parsed once into a domain type.

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.