agentleFS
devlintdevlint/GitWand/.claude/skills/add-resolver/SKILL.md

add-resolver

Skill194 starsChanged 5 months agoMITView on GitHub

Add a format-specific resolver to GitWand core. Use this skill whenever someone wants to add support for a new file format, file extension, lockfile, config file type, or format-specific conflict resolution strategy. Triggers: "add resolver for .prisma", "handle composer.lock conflicts", "resolve .tf files", "new format support", "add lockfile resolver", "teach GitWand about X files".

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    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:

    Is it a lockfile?: If yes → confidence is 0.95 by convention.

    line 25
  2. Very specific:

    Create packages/core/src/resolvers/<format>.ts.

    line 49
  3. Very specific:

    Create packages/core/src/__tests__/resolvers/<format>.test.ts with at least

    line 146
  4. Distinctive:

    If the resolver touches the main resolution pipeline, also run the parity probe to make sure Rust and TypeScript outputs stay in sync:

    line 192

≈ “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

Skill: add-resolver

Guide the agent through adding a format-specific resolver to packages/core end to end — file creation, dispatcher wiring, tests, and docs — without missing any of the four required touch points.


Step 1 — Qualify the need

Answer these before writing any code:

  1. Format / extension? (e.g. .prisma, composer.lock, .tf, .toml)
  2. Is it a lockfile? If yes → confidence is 0.95 by convention. Lockfile conflicts are almost always trivially re-generable, so high confidence is appropriate.
  3. Resolution strategy? Examples:
    • Structured config (key = value) → merge non-conflicting keys
    • Lockfile → accept one side; file is regenerable
    • Source code → be conservative, 0.70–0.80 max
  4. Browser-compatible parsing available? packages/core must run in browsers — no Node.js imports (fs, path, child_process). If no parser lib is available, write a pure string-manipulation parser.

Confidence score guidelines:

File typeScore
Lockfile0.95
Structured config (JSON, YAML, TOML, dotenv)0.85–0.90
Source code0.70–0.80
Never exceed 0.95 unless perfectly deterministic—

Step 2 — Create the resolver file

Create packages/core/src/resolvers/<format>.ts. Follow this exact shape, which mirrors all existing resolvers in the directory:

/**
 * GitWand — <FORMAT> resolver
 *
 * Strategy: [describe the resolution strategy precisely]
 *
 * Handled cases:
 *  - [case 1]
 *  - [case 2]
 *  - [non-resolvable case → return null]
 */

// ✅ Allowed: pure string manipulation, parsing, algorithms
// ❌ FORBIDDEN: import * as fs from 'fs' / import * as path from 'path'

export interface ResolveResult {
  /** Resolved lines, or null if not resolvable */
  lines: string[] | null;
  /** Human-readable description of what was done (or why it failed) */
  reason: string;
}

/**
 * Attempts to resolve a <format> conflict.
 * Confidence: 0.XX — [justify the score here]
 */
export function tryResolve<Format>Conflict(
  baseLines: string[],
  oursLines: string[],
  theirsLines: string[],
): ResolveResult {
  // ... resolution logic ...

  return { lines: null, reason: "Non-resolvable: [reason]." };
}

Key points:

  • Zero Node.js imports — double-check every import before writing it
  • Justify the confidence score in the JSDoc comment
  • On parse failure, return lines: null rather than throwing — the dispatcher falls back gracefully when a resolver returns null

Step 3 — Wire into the dispatcher

Open packages/core/src/resolvers/dispatcher.ts and make four changes:

3a. Add a detection function in the // ─── Type detection ─── section:

/** Checks whether the file is a <format> file */
export function is<Format>File(filePath: string): boolean {
  return /\.<ext>$/i.test(filePath);
  // For exact filenames: /(?:^|[\\/])<name>$/i.test(filePath)
}

3b. Add the import at the top with the other resolver imports:

import { tryResolve<Format>Conflict } from "./<format>.js";

3c. Add the route in tryFormatAwareResolve(), keeping specificity order — more specific detections first (e.g. composer.lock before *.lock, a named lockfile before a generic JSON):

if (is<Format>File(filePath)) {
  const result = tryResolve<Format>Conflict(
    hunk.baseLines,
    hunk.oursLines,
    hunk.theirsLines,
  );
  return {
    lines: result.lines,
    reason: `[<format>] ${result.reason}`,
    resolverUsed: "<format>",
  };
}

3d. Extend the resolverUsed union in FormatResolveResult:

resolverUsed: "json" | "markdown" | ... | "<format>" | "structural" | "none";

Step 4 — Write tests

Create packages/core/src/__tests__/resolvers/<format>.test.ts with at least 5 cases:

import { describe, it, expect } from "vitest";
import { resolve } from "../../resolver.js";

// Case 1 — Simple resolvable conflict (happy path)
describe("<FORMAT> — non-conflicting addition", () => {
  it("auto-resolves and keeps both sides", () => { /* ... */ });
  it("reason mentions [<format>]", () => { /* ... */ });
});

// Case 2 — Irresolvable complex conflict
describe("<FORMAT> — irresolvable conflict", () => {
  it("returns resolved: false", () => { /* ... */ });
});

// Case 3 — Empty / minimal structure
describe("<FORMAT> — minimal file", () => {
  it("does not throw", () => { /* ... */ });
});

// Cases 4 & 5 — Representative real-world conflicts for this format
describe("<FORMAT> — real scenario 1", () => { /* ... */ });
describe("<FORMAT> — real scenario 2", () => { /* ... */ });

Build fixtures using real git conflict markers (<<<<<<< ours / ||||||| base / ======= / >>>>>>> theirs). Never mock the diff algorithms.


Step 5 — Update the docs

In packages/core/CLAUDE.md, add the new resolver name to the resolvers list.


Step 6 — Validate

cd packages/core && pnpm test

If the resolver touches the main resolution pipeline, also run the parity probe to make sure Rust and TypeScript outputs stay in sync:

cd apps/desktop && pnpm test:parity

Common mistakes to avoid

  • Forgetting step 3d — omitting the new name from the resolverUsed union causes a TypeScript error that only surfaces at build time.
  • Wrong specificity order — a generic *.lock matcher placed before pnpm-lock.yaml would silently shadow the existing pnpm resolver.
  • Importing Node.js builtins — packages/core runs in browsers and Tauri WebViews; any Node.js import will crash at runtime in those environments.
  • Throwing on parse errors — always return lines: null with a reason; let the dispatcher fall through to the text engine.

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.