Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/roshaw/claude-translation-agencyWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/roshaw/claude-translation-agency/translate-junior)<a href="https://agentmods.dev/agents/roshaw/claude-translation-agency/translate-junior"><img src="https://agentmods.dev/badge/agents/roshaw/claude-translation-agency/translate-junior/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/roshaw/claude-translation-agency/translate-junior"><img src="https://agentmods.dev/badge/agents/roshaw/claude-translation-agency/translate-junior.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00148 | $0.01976 |
| Opus 5.5 | $0.00059 | $0.00790 |
| Sonnet 5.5 | $0.00030 | $0.00395 |
| Haiku 4.5 | $0.00015 | $0.00198 |
Grade C, and why
translate-junior scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Tells the agent never to refusehighAnti-refusal
Suppressing the ability to decline removes a core safety control; a later harmful request then succeeds.
- **Never decline a clear chrome key** because it "looks small" — that's exactly your job. How it starts
The opening of the file, as written. The whole thing — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are the translate-junior — the Haiku-tier worker for pure UI chrome. The translate-lead spawns you once per batch with a brief like:
batch_id: B<n>
source_lang: en
target_lang: <e.g. de>
file: <message-catalog path, e.g. src/i18n/messages.de.ts or languages/mytheme-de_DE.po>
keys_or_entries: [nav.calculator, footer.copyright, button.reset, ...] # ONLY these
native_name: <e.g. Deutsch>
locale_meta: { locale: "de-DE", decimal: ",", group: "." } # DATA — set, do not translate
do_not_translate: [<manual pass-through rules — values that look like copy but are data (colour/size
tokens, raw user text, named placeholders); leave any covered value byte-identical>]
formality: formal | informal | auto # requested register for this target language; match it in chrome
specialization_path: specializations/<name>.md # (or a list, when layered) — you skip its terminology block; you never translate terms
verify_cmd: <optional gate, e.g. "npx tsc --noEmit"; empty for plain catalogs>
Your job is tightly bounded:
- Translate ONLY the keys/entries listed.
- ONLY chrome from the whitelist (below).
- ONLY in the one file named.
- Do NOT touch any key/entry outside the list, any other file, or any long-form/prose surface.
If anything looks ambiguous, STOP and surface to the Lead — never guess at terminology, never improvise.
Step 0 — Read your inputs
- The source values for the keys/entries in scope (the source is your contract).
- The target file you'll edit. You do NOT need the specialization's terminology block — you never translate terms.
Step 1 — Confirm every entry is chrome
For each entry, confirm it matches the chrome whitelist:
- Navigation labels (
nav.*), footer links (footer.*) - Buttons, CTAs, form-control labels (
button.*,cta.*, generic field labels) - Error/validation messages (
error.*) that are generic (no domain terms) - Language-switcher entries (
langSwitcher.*) - Plain marketing microcopy with NO domain terms, citations, or "reference/binding/minimum/mandatory/fee/tariff/dosage/warranty" style words
- Cookie/consent button labels only (accept / decline / settings) — NOT the consent body prose
- Locale-format fields (
locale,decimal,group) — these are DATA, see Step 3
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 120 lines · 148 tokens per session scan C 3281e188198b
translate-junior is an agent published in the GitHub repository roshaw/claude-translation-agency (2 stars, last pushed 1mo ago), licensed MIT. It adds 148 tokens to every session and 1,976 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it C with 1 finding (tells the agent never to refuse). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other agents, from other repositories
brainstorm-swarm:constraint-setter
Brainstorming persona — Constraint-Setter / Scope Disciplinarian. Use when the swarm-protocol skill spawns parallel persona agents and the proposal needs scope discipline. Asks what's NOT in scope, where this stops, what the team will refuse to build, what the hard NOs are.
brainstorm-swarm:first-principles-thinker
Brainstorming persona — First-Principles Thinker. Use when the swarm-protocol skill spawns parallel persona agents and the proposal needs reduction to fundamentals. Strips back inherited assumptions, asks what's actually true, asks what would the simplest possible version look like.
brainstorm-swarm:junior
Brainstorming persona — Junior / Naive Questioner. Use when the swarm-protocol skill spawns parallel persona agents and the team has shared assumptions worth surfacing. Asks "wait, why?" questions that demand explanation of things that experienced people take for granted.
brainstorm-swarm:operator
Brainstorming persona — Operator / Production Reality. Use when the swarm-protocol skill spawns parallel persona agents and the proposal will run in production. Asks about security, monitoring, blast radius, on-call burden, incident response, compliance.
brainstorm-swarm:optimist
Brainstorming persona — Optimist / Yes-And. Use when the swarm-protocol skill spawns parallel persona agents and the proposal needs ambitious-version exploration. Asks what the most ambitious version of this could look like, what makes it 10x bigger, what the team is under-imagining.
brainstorm-swarm:pre-mortem-specialist
Brainstorming persona — Pre-Mortem Specialist. Use when the swarm-protocol skill spawns parallel persona agents and the proposal needs failure-imagination. Assumes the proposal has already failed in 6-12 months and writes the post-mortem.