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.
npx agentmods add commands/deepfusionlabs/deep-init/translategit clone --depth 1 https://github.com/deepfusionlabs/deep-initWhat 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 | $0.00035 | $0.00402 |
| Opus 5 | $0.00017 | $0.00201 |
| Sonnet 5 | $0.00007 | $0.00080 |
| Haiku 4.5 | $0.00003 | $0.00040 |
Grade A, and why
translate scanned grade A with 0 findings 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 2d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Run the deep-init skill in translate mode. Load skills/deep-init/SKILL.md and follow references/i18n.md (read references/report.md first).
If no language is given in $ARGUMENTS, open the language picker — the Translate picker flow in SKILL.md, via the AskUserQuestion tool (buttons, no flag to type): the shipped targets (Spanish · Hebrew) as buttons + Other; choosing Other accepts any language you type (the other: <language> escape hatch — content translated, chrome falls back to English). Map the choice to a --translate=<lang> code.
Then:
- Build/refresh the canonical English report first (
report.html) — it is never altered. - Run the content translation pass over the report's prose fields into
.ai/i18n/translation_memory.json(mask grounded tokens — code,file:line, record IDs, product nouns — translate, restore, verify; honest-degrade any miss to English). - Emit
report.<lang>.htmlviapython tools/build_i18n.py <dir> --lang <code>(deterministic; sets<html lang dir>, RTL for Hebrew; chrome is the template's bakedSTRINGStable picked by<html lang>).
English is always generated as the canonical output; translation is a derived overlay and never changes a grounded claim. Shipped languages: es he — any other language works on demand via other: <language> (chrome falls back to English, stated).
$ARGUMENTS
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.
- 2d ago First seen · 18 lines · 35 tokens per session scan A 11765ef4f9de
translate is a command published in the GitHub repository deepfusionlabs/deep-init (6 stars, last pushed 13d ago), licensed MIT. It adds 35 tokens to every session and 402 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
claude-md-migrate
Rewrite a bloated or stale CLAUDE.md into a lean, verified "map, not wishes" file — nothing invented, hard rules preserved verbatim.
fleet-conformance
Scan every repo on the machine for guardrails, testing, and observability conformance; audit the deltas semantically; produce a fleet report; and propose canary-first remediation.
claude-md-new
Scaffold a CLAUDE.md for this repo from battle-tested templates, filled in with the project's real commands.
implement-review
One agent implements a task, then reviews its own work behind a hard verification gate before finalizing.
claude-md-audit
Grade this repo's CLAUDE.md / AGENTS.md (0–100) and return a worst-first fix list.
project-init
Stand up the context layers for a project — a thin pointer-style AGENTS.md plus seed compass maps — without touching the memory layer.