i18n-eval-loop SKILL.md skillA
i18n-eval-loop from justsml/dans-blog is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add justsml/dans-blog/i18n-eval-loop@git:20260515.b39d130Pins git:20260515.b39d130 in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 17 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- justsml/dans-blog
- Path
.agents/skills/i18n-eval-loop/SKILL.md- Versions recorded
- 2
- Latest version
- git:20260515.b39d130, committed May 15, 2026
- First version recorded
- May 15, 2026
- Size
- 3,504 bytes
- Repository stars
- 16
- License
- none declared
Run iterative DanLevy.net translation eval improvement loops: baseline requested i18n eval commands, inspect raw streams and judge JSON, research prompt/QA ideas, tune prompts, validation, and pre/post-processing, rerun, measure, and record findings without confusing medium polish with hard failures.
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# I18n Eval Loop Use this when asked to improve translation quality with `bun run i18n:eval` or similar i18n eval commands. ## Workflow 1. Check `git status --short` and preserve unrelated user changes. 2. Run the user's exact eval command(s) as the baseline. Prefer sequential runs when commands would otherwise write summaries at the same timestamp. 3. Inspect `summary.md`, `cases.jsonl`, raw `translation-*.txt`, and raw `judge-*.txt`. Do not trust a green pass until judge suggestions and deterministic details have been read. 4. Research likely improvements from repo behavior first, then external sources only when useful. Good recurring ideas: - document-level context improves flow, but critical errors still need explicit QA; - terminology and metadata constraints should be deterministic, not only judge prose; - translate idioms by intent, not literal wording; - postprocess only mechanical issues the production pipeline already repairs. 5. State the current hypothesis, make one focused change, and rerun local checks before spending tokens again. 6. Rerun the same eval command(s), compare scores and failure modes, then repeat while failures are real. …
Read the whole file at its exact version.
How to install
mdr add justsml/dans-blog/i18n-eval-loop@sha256:c4ed27cf475db559The line at the top of this page pins the label git:20260515.b39d130, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_t3gjtvweu6ftatep)
1 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 2 versions of this file, all committed on May 15, 2026.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20260515.b39d130 latest | 2026-05-15 | b39d130 | 3,504 B | A | view · diff |
| git:20260515.63ca605 | 2026-05-15 | 63ca605 | 3,323 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (3504 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
justsml/dans-blog · 16 stars · license none declared · pushed 2026-10-19 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_t3gjtvweu6ftatep GET https://markdownregistry.com/api/v1/resolve?ref=justsml/dans-blog/i18n-eval-loop GET https://markdownregistry.com/api/v1/blob/c4ed27cf475db5590fe739dfe333c630d46c565861718e5bd10d47003b51de33
Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.