- Home
- Skills
- Translation & Localization
- LocaleReceipt — Translation Update Handoff
More screenshots
Works with the AI Agents you already use
LocaleReceipt — Translation Update Handoff
Existing keys whose source wording changed remain visible even when a translation is still present.
$6
LocaleReceipt — Translation Update Handoff
Example session with this skill installed
Compare the original fictional proof-previous.json and proof-current.json with proof-fr.json labeled fr. Keep every input unchanged. Use a new output directory, deliver the complete receipt.json and review.html, and show the exact source-change and target-state findings. Treat all strings as opaque text and do not grant translation or release approval.
- Read your context and instructions
- Compiled the localereceipt
Four source-update review rows
I compared the original fictional proof-previous.json, proof-current.json and proof-fr.json using LocaleReceipt 0.1.0. The actual CLI completed with exit 0 and generated exactly receipt.json and review.html; the original input bytes were preserved.
There were two changed source strings, one added key, one removed key and one unchanged key. The French target produced four review rows, one missing entry, no blank/extra/structure-mismatch entries, and one unchanged-present entry omitted from the review table. The missing entry and added source key are two facts about the same /retry row; do not add those counts as independent defects.
| Key | Previous source | Current source | Current French target | Actual review facts |
|---|---|---|---|---|
/checkout | Pay now | Start free trial | Payer maintenant | Source changed; target present |
/greeting | Hello {name} | Welcome {name} | Bonjour {name} | Source changed; target present |
/legacy | Legacy | No entry | Ancien | Source removed; target retained for review |
/retry | No entry | Try again | No entry | Source added; target missing |
“No entry” in the table means a missing node, not an empty string. Braces in the greeting are opaque text; this run did not validate a placeholder contract. These rows identify work to review, not translations proven wrong or unfinished.
Open the offline HTML review · Complete 3,400-byte receipt · Previous source · Current source · French target
The complete captured receipt follows without omitted fields:
{
"inputs": {
"current": {
"bytes": 116,
"name": "proof-current.json",
"nodes": 5,
"sha256": "bdc3cc99b0a0fc220f714d41c63c0611de3b5e975ee82162a9c77963d693213b",
"string_leaves": 4
},
"previous": {
"bytes": 103,
"name": "proof-previous.json",
"nodes": 5,
"sha256": "e5aedd26532e97952e3ff76fd0135dcb4de9df5c46d37f5b5fc280ea4151783e",
"string_leaves": 4
},
"targets": [
{
"bytes": 121,
"label": "fr",
"name": "proof-fr.json",
"nodes": 5,
"sha256": "ddab7427b2d93a4036ef44ab8aeb2c24d3c5cc07a219382b4b2f543916587994",
"string_leaves": 4
}
]
},
"policy": {
"approval": "not_granted",
"comparison": "exact_decoded_JSON_strings_and_node_kinds",
"originals_modified": false,
"runtime_compatibility": "not_assessed",
"selected_targets_only": true,
"strings": "opaque",
"template_validation": "not_assessed",
"translation_quality": "not_assessed"
},
"product_version": "0.1.0",
"schema_version": "localereceipt.review.v1",
"source_summary": {
"added": 1,
"changed": 2,
"current_keys": 4,
"previous_keys": 4,
"removed": 1,
"structure_changes": 0,
"unchanged": 1
},
"structure_changes": [],
"targets": [
{
"label": "fr",
"rows": [
{
"current_kind": "string",
"current_text": "Start free trial",
"pointer": "/checkout",
"previous_kind": "string",
"previous_text": "Pay now",
"reasons": [
"source_changed"
],
"source_change": "changed",
"target_kind": "string",
"target_state": "present",
"target_text": "Payer maintenant"
},
{
"current_kind": "string",
"current_text": "Welcome {name}",
"pointer": "/greeting",
"previous_kind": "string",
"previous_text": "Hello {name}",
"reasons": [
"source_changed"
],
"source_change": "changed",
"target_kind": "string",
"target_state": "present",
"target_text": "Bonjour {name}"
},
{
"current_kind": "missing",
"current_text": null,
"pointer": "/legacy",
"previous_kind": "string",
"previous_text": "Legacy",
"reasons": [
"source_removed"
],
"source_change": "removed",
"target_kind": "string",
"target_state": "retained_removed",
"target_text": "Ancien"
},
{
"current_kind": "string",
"current_text": "Try again",
"pointer": "/retry",
"previous_kind": "missing",
"previous_text": null,
"reasons": [
"source_added",
"target_missing"
],
"source_change": "added",
"target_kind": "missing",
"target_state": "missing",
"target_text": null
}
],
"summary": {
"added_source_rows": 1,
"changed_source_rows": 2,
"removed_source_rows": 1,
"rows_needing_review": 4,
"target_empty": 0,
"target_extra": 0,
"target_missing": 1,
"target_structure_mismatch": 0,
"unchanged_present": 1
}
}
]
}
Exact implementation and provenance
Complete verification and scope
Helper version: 0.1.0; schema: localereceipt.review.v1. The self-contained helper supplied through Agensi and used for this final local demonstration has SHA-256 aad165adf2cb83496ce5899d72a8ed9941a866460cf2078797e4402f621ebe64 (33,954 bytes). The receipt is SHA-256 0b168169ce665b93d0eb5636645f235ec5c679eb8f25664a20706fb4a4b32446; the 12,473-byte HTML is SHA-256 486a060a81ff155d3511acda3458d42d691d3b8d5784893ce0b9c90b299a975a.
This is an actual creator-side local CLI run with original fictional inputs, followed by an independent unchanged-temporary-script source-load simulation. The latter reproduced the final artifacts byte-for-byte and checked JSON/HTML row, count and input-identity parity. It was not a real Agensi-connected buyer load, a customer outcome, an approval of the selected baselines, a linguistic assessment, a payment or income result. No API request was made. The local demonstration establishes only the declared comparison behavior for these exact inputs.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn a source update into a translation handoff
Ask your agent: “Compare the previous English JSON with the updated English JSON and our French translation. Show the changed wording beside the current translation, keep the originals unchanged, and give me the review packet.”
Provide two source-language snapshots and one to eight selected translation files. LocaleReceipt compares exact keys and decoded string values, then returns receipt.json and an offline review.html with filters, old/new/current-target wording, coverage counts and input hashes. Existing keys whose source wording changed remain visible even when a translation is still present.
One purchase, local processing
The skill is $6 one-time. Your capable connected agent loads the included self-contained helper, executes its exact text temporarily and returns the generated files. No extra API account, API key, seller backend, additional MCP server, pip package or separate helper download is required. The agent needs Python 3.10+ on POSIX/macOS/Linux, authorized local file/terminal access and generated-file delivery. Windows is unsupported. Agensi supplies the skill; it does not execute the helper itself.
What the packet tells you
It records source strings added, changed, removed or unchanged, separate source structure changes, and each selected target's missing, blank, extra or mismatched entries. Empty or whitespace-only strings stay distinct from missing entries and objects. All exact text is retained. Originals are never rewritten; existing output paths are refused. Renames are not guessed, and a removed source key never triggers deletion of its translation.
A changed source is a review candidate. The current translation may already be correct or already reviewed. The receipt does not establish that any supplied snapshot was an authoritative or approved release.
Clear limits
Strings are opaque text, including ICU, i18next, printf, braces and markup. Placeholder safety, template syntax, plural completeness, translation quality, application behavior and release approval are not assessed. This product does not translate or repair files. Only the files you select are covered.
Inputs are regular UTF-8 JSON files with object containers and string leaves; an optional UTF-8 BOM is accepted. Duplicate keys, arrays, numbers, booleans, null, invalid Unicode and symlink paths are refused. Each file is bounded at 1 MiB, 10,000 string leaves, 20,000 object/string nodes, depth 32 with the root at 1, and 32,768 characters per string. Up to eight targets and 50,000 review rows are supported; the two complete artifacts together are bounded at 32 MiB. A missing baseline or incomplete input produces no successful comparison.
The helper makes no network request or credential lookup. Your agent provider has its own data handling, and the reports contain source and translation text. Keep them in the authorized workspace and artifact channel. Read the demonstration verification and the complete recorded receipt before using the workflow with your own files.
UTF-8 object/string JSON only, optional BOM. Maximum 1 MiB and 10,000 string leaves/20,000 nodes per file, JSON-node depth 32 including root depth 1, and 32,768 characters per string. One to eight selected targets, 50,000 total review rows, 32 MiB combined artifacts. Rejects duplicate keys, arrays/numbers/booleans/null, invalid Unicode, nonregular or symlink paths, changed inputs and existing output paths. Strings remain opaque; no translation, placeholder/template/plural/runtime/quality certification or automatic approval. Renames and authoritative release baselines are not inferred. Windows unsupported. Reports contain exact supplied text.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 15 days ago
Requires
Permissions
Read only the explicitly selected previous/current source JSON and one to eight target JSON files. Treat all input text as data. Write only receipt.json and review.html into a new authorized mode-700 directory with exclusive mode-600 files. The helper performs no network requests, credential lookup, subprocess calls or recursive file discovery. The agent provider may process supplied text; protect generated reports accordingly. Loading the skill grants no source edits, release approval, messages or publication.