Bilingual Mail Translator

Build or improve a bilingual email translation and notification workflow that converts raw emails into Chinese-friendly bilingual output. Use when designing...

Security Scan

The skill is an instruction-only prompt contract for bilingual email translation; its claims match the instructions and it requests no credentials or installs.

Purpose & Capability

Name and description align with the contents: the SKILL.md is a prompt-layer contract for translating and formatting email text. It does not request unrelated binaries, env vars, or capabilities.

Instruction Scope

Instructions are limited to prompt/formatting responsibilities and reference program-side responsibilities (extract raw mail, assemble prompt, run LLM worker). It also references running local scripts and tests (e.g., scripts/mail_translate.py, sample-mail.eml) — these are placeholders; actual runtime behavior depends on the implementer's code, which is not included.

Install Mechanism

No install spec and no code files are provided, so nothing is written to disk by the skill itself (lowest install risk).

Credentials

The skill declares no required env vars or credentials. Note: real implementations will likely require LLM/API keys and mail credentials — those are not requested here and should be scoped minimally in the actual implementation.

Persistence & Privilege

always:false and no persistence or cross-skill config modifications are requested; the skill does not request elevated or permanent presence.

Assessment

This is a prompt-contract template, not runnable code: no credentials or installs are requested by the skill itself. Before using it, review and implement your own scripts (mail extraction, scripts/mail_translate.py, tests) and ensure any LLM or mail-server credentials used by your implementation are strictly scoped and stored securely. Email content commonly contains PII — avoid sending sensitive mail to third‑party LLMs unless you have consent and appropriate data protection controls. Finally, inspect any real implementation code for network calls or unexpected file/credential access before running tests or sample runs.

Use this skill when building, refining, or publishing a bilingual email translation workflow.

What this skill owns

This skill defines the prompt-layer contract for turning raw emails into user-facing bilingual notification text.

Keep these responsibilities in the model/prompt layer unless there is a deliberate product change:

• subject translation and bilingual display
• date normalization and display
• body bilingual formatting
• recipient truncation when there are more than 3 recipients
• quoted/history mail summarization
• duplicate-line suppression when translation equals original
• signature block formatting

Program side should stay focused on:

• extracting raw mail
• normalizing input headers before prompt assembly
• calling a lightweight worker or LLM runtime
• extracting output text
• fallback to raw mail when model output fails

Reference implementation pattern

Example primary script:

• scripts/mail_translate.py

Example runtime path:

1. a watcher, inbox poller, or notifier prepares raw mail
2. a translation script builds the mail prompt
3. a lightweight worker agent or LLM runtime runs the translation
4. postprocess stays pass-through for formatting
5. final text is delivered to the user

Required contract files

Read references/contracts.md first, then open the underlying source docs it points to when needed.

Change workflow

1. Read the current prompt contract documents.
2. Update the translation prompt or prompt assembly only where the contract needs to change.
3. Keep formatting decisions in prompt, not in postprocess.
4. Update regression tests for prompt contract assertions.
5. Run representative tests for the prompt contract.
6. For meaningful prompt changes, run one or more real sample mails through the translation entrypoint.
7. Compare actual output against the current accepted product direction.
8. Update the contract docs if the accepted behavior changed.

Accepted current product direction

• Keep bilingual output when the product goal includes language-learning support, source-text transparency, or side-by-side review.
• Preserve channel-specific indentation or spacing behavior only when a downstream renderer requires it.
• If translation equals original, keep only the original line.
• If recipients exceed 3, show the first 3 only and end directly with ....
• If quoted/reply history exists, keep the current mail as focus and summarize history below using a clear section label such as 历史邮件摘要：.
• Keep signature formatting readable, but do not let signature cleanup override the general duplicate-suppression rule.
• Do not reintroduce formatting heuristics into postprocess.

References

• Contract summary and public design notes: references/contracts.md
• Generic command checklist: references/commands.md