adopt-better-result — Claude Code skill: error-handling audit
This skill audits error handling in an existing TypeScript codebase and proposes a move to TaggedError and Result.
npx skills add dmmulroy/better-result --skill adopt-better-result -a claude-codeWhat adopt-better-result does
The skill runs in two modes. The audit branch reads repository instructions, package manifests, source layout, test commands and architecture docs, then checks whether better-result is installed and which version of the API is available. It maps every production-code area, skipping generated and vendored code, and traces search hits into end-to-end failure paths. The report goes to the project's usual documentation location, or to docs/better-result-adoption.md if there isn't one.
The named-slice branch handles implementation. It kicks in when you explicitly ask to rewrite a function, module or boundary, or when you approve a slice from the audit. Evidence collection stays inside that slice and its propagation path, and the path ends up consistently Result-based from failure source to the chosen handling boundary. Format, lint, type-check and relevant tests run at the end.
One thing to know before you start: the audit branch makes no production-code changes. The report is the only thing it writes, and implementation waits for your explicit approval of one named slice. Expecting edits right after the first run will leave you disappointed.
What it helps with
- audit error handling across a whole repository
- get a report proposing a Result migration
- rewrite one module with TaggedError
- pick the first vertical slice to migrate
- confirm a failure path is covered to its handling boundary
Example prompts for the agent
- Audit error handling in this repository and propose a better-result migration plan.
- Implement the named slice: the payments module from failure source to handling boundary.
- I approve the first slice from the report — start the implementation.
How to install adopt-better-result
1Skills CLI (Vercel)
npx skills add dmmulroy/better-result --skill adopt-better-result -a claude-code2Manually: put the files in place
~/.claude/skills/adopt-better-result/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/adopt-better-result/
1Skills CLI (Vercel)
npx skills add dmmulroy/better-result --skill adopt-better-result -a codex2Manually: put the files in place
.agents/skills/adopt-better-result/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add dmmulroy/better-result --skill adopt-better-result -a cursor2Manually: put the files in place
.cursor/skills/adopt-better-result/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install dmmulroy/better-result adopt-better-result2Manually: put the files in place
.github/skills/adopt-better-result/SKILL.md1Skills CLI (Vercel)
npx skills add dmmulroy/better-result --skill adopt-better-result -a gemini-cli2Platform CLI
gemini skills install https://github.com/dmmulroy/better-result --path skills/adopt-better-resultManually: put the files in place
.devin/skills/adopt-better-result/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add dmmulroy/better-result --skill adopt-better-result -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about adopt-better-result
How do I install adopt-better-result in Claude?
Run npx skills add dmmulroy/better-result --skill adopt-better-result -a claude-code. The skill lands in ~/.claude/skills/adopt-better-result/SKILL.md.
What does adopt-better-result do?
It audits error handling in an existing TypeScript codebase and proposes a move to TaggedError and Result. It can also implement one named vertical slice.
Does the skill change code during the audit?
No. On the audit branch the report is the only repository change. Implementation starts only after you approve one named slice.
Source text
SKILL.md · MIT
--- name: adopt-better-result description: Adopt better-result in an existing TypeScript codebase. Use for a repository-wide error-handling audit and proposal, or for implementing a named vertical slice with TaggedError and Result. --- # Adopt better-result Start with evidence, then work at the scale named by the user. ## Choose the branch - **Audit branch:** Use when the user requests a repository-wide audit or proposal, or has not named an implementation slice. Complete the audit branch before changing production code. - **Named-slice branch:** Use when the user explicitly requests implementation of a function, module, boundary, or previously proposed slice. Limit evidence collection to that slice and its propagation path. An approved slice from the audit branch continues through the named-slice branch. ## Audit branch: repository-wide proposal ### 1. Establish the repository and API facts Read the repository instructions, package manifests, source layout, test commands, and architecture documentation. Determine whether `better-result` is installed and which version/API is available. Prefer installed types and source over remembered APIs. Map every production-code area before studying individual failures. Exclude generated and vendored code from the audit; inspect tests for intended behavior. **Complete when:** every production-code area is listed, the validation commands are known, and the available `better-result` API has been verified. ### 2. Audit every production-code area Follow [`references/repository-audit.md`](references/repository-audit.md). Trace search evidence into end-to-end failure paths, keep grouped sites accountable with counts, and record obscured paths as unknown. Write the audit to the repository's established documentation or planning location, falling back to `docs/better-result-adoption.md`. **Complete when:** the report satisfies the coverage and completion requirements in `repository-audit.md`. ### 3. Design the target model Apply the failure taxonomy in [`references/tagged-errors.md`](references/tagged-errors.md) and make each boundary decision with [`references/result-boundaries.md`](references/result-boundaries.md). Add the proposed dispositions, user-facing mappings, boundary decisions, and ranked vertical migration slices to the report. Recommend the best first slice and explain why it is bounded and valuable. **Complete when:** every catalogued failure has a disposition, every boundary has a recorded codec decision, and one first slice is recommended. ### 4. Present the approval gate Present the report and recommendation. The audit branch's only repository change is the report. **Complete when:** the report is presented. An explicit approval of one named vertical slice starts the implementation branch; otherwise the run ends here. ## Named-slice branch: implementation ### 1. Establish the slice and API facts Read the repository instructions, validation commands, installed `better-result` API, relevant production code, and tests. Trace the named failure source through its immediate callers to its handling or transport boundary. Reuse an existing adoption report when present; otherwise retain the evidence for the implementation summary. **Complete when:** the slice boundary, current behavior, failure variants, callers, side effects, and transport boundaries are accounted for, and the available API is verified. ### 2. Implement one vertical slice Follow [`references/vertical-slice-migration.md`](references/vertical-slice-migration.md), using [`references/tagged-errors.md`](references/tagged-errors.md) for failure dispositions and [`references/result-boundaries.md`](references/result-boundaries.md) for boundary contracts. Keep the implementation within the named propagation path. **Complete when:** the path is consistently Result-based from failure source to its chosen handling boundary, with every error variant handled and every control-flow seam accounted for. ### 3. Verify and report Run the repository's applicable format, lint, type-check, and relevant test commands. Update an existing adoption report with completed work, deviations, discoveries, and validation evidence; otherwise include them in the user-facing summary. End after recommending the next slice. **Complete when:** all applicable checks pass, or every failure is reported with evidence, and the report or summary accurately describes the repository's current state.
© dmmulroy · MIT · dmmulroy/better-result