How to use the Fill Translations agent skill

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file of instructions that an AI agent loads when a request matches the skill's description. This skill is named fill-translations and is designed for the Reactive Resume repository.

When it runs, the agent extracts text strings from both apps, finds active catalog entries that are missing translations across every configured locale, and fills them in. It then runs Lingui checks to confirm nothing is missing and that placeholders still match. It is built to preserve existing translations and does not claim the translations are linguistically perfect.

• Finds missing entries in web and mobile locale catalogs, including older gaps unrelated to the current task unless you narrow the scope.
• Splits work across parallel helper agents when delegated-task or subagent support is available, with a sequential fallback.
• Verifies completeness with Lingui checks and compile validation before reporting.

Install the skill (ZIP download or Claude apps)

Download the skill ZIP from this page and unzip it. The skill folder is named fill-translations and contains SKILL.md, agents/openai.yaml and evals/evals.json.

For a personal install, unzip it into ~/.claude/skills/. To share it with a single project, unzip it into the project's .claude/skills/ folder instead.

You can also upload the skill in the Claude apps under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills

Requirements before you start

The skill runs from the root of a Reactive Resume checkout. It follows that repository's AGENTS.md, command wrappers and current package scripts, and it does not need a translation service or a new dependency.

You need pnpm and the project's workspace dependencies, which include the Lingui 6 tooling. For parallel work you need delegated-task or subagent support in your agent environment; without it the agent completes the same inventory sequentially and tells you so.

• A checkout of the Reactive Resume repository
• pnpm and the project's workspace dependencies (Lingui 6 tooling)
• Delegated-task or subagent support for parallel work (sequential fallback exists)

Trigger the skill with natural requests

The agent loads the skill when your request matches its description. Phrases that match include fill translations, complete translations, missing locale strings, and a request to make sure translations are complete before a release.

Be specific about scope. By default the skill covers existing gaps outside your current task. If you only want part of the work, say so explicitly, because the skill will otherwise cover everything it finds.

• "Fill translations for the Reactive Resume repo before the release."
• "Complete the missing locale strings in both web and mobile."
• "Make sure translations are complete before we ship."
• "Only fill the missing mobile locale strings, not web."

Safety notes: scripts and network

The skill runs scripts: it executes pnpm commands such as Lingui extraction, checks and compilation, and it edits catalog files. The source states that it does not use the network.

It is not a read-only audit. The skill is designed to write to catalog files and is told not to activate when edits are forbidden. Compilation writes locale .js files, and the workflow restores pre-existing outputs and removes only files it created. Commit and review changes yourself; the skill is not instructed to commit, push or release unless you ask.

• Runs scripts: yes. Uses network: no.
• Edits catalog files, so do not use it for a read-only status check.
• Keep your staged and unstaged changes; the skill does not reset, checkout or clean the checkout.

Review the results and what the skill does not prove

The skill checks that catalogs are technically complete: entries are filled before fallback, placeholders and tags match, and the whole catalog passes sync, missing and compile checks. Technical completeness and linguistic quality are separate claims.

Review the diff yourself, especially glossary-sensitive wording and translations that are identical to the English source. Native-speaker review remains the right step for quality.

• Run git diff to read the added translations.
• Check glossary terms against GLOSSARY.md in the repository.
• Treat a passing check as proof of completeness, not of translation quality.