Email Triager

Triage, categorize, and draft responses to emails. Sorts by urgency, flags action items, and generates context-aware reply drafts.

Security Scan

The skill's requirements and instructions are consistent with an email-triage assistant; it is instruction-only, asks for no credentials or installs, and its behavior matches the description.

Purpose & Capability

Name and description (triage, categorize, draft replies) align with the SKILL.md and README. The skill is instruction-only and requests no unrelated binaries, env vars, or credentials, which is appropriate for a text-based email triager.

Instruction Scope

Runtime instructions operate on email text provided by the user (batch or single). They do not instruct the agent to access system files, environment variables, or external endpoints for processing. They do reference an optional external 'context pack' URL for domain-specific handling; otherwise the instructions stay within the stated purpose.

Install Mechanism

No install spec or code is included (instruction-only). README shows a manual copy/install hint, which is normal. Nothing is downloaded or written to disk by the skill itself.

Credentials

The skill declares no required environment variables, credentials, or config paths. That is proportionate for a tool that processes user-provided email text.

Persistence & Privilege

No 'always: true' or other elevated persistence is requested. The skill is user-invocable and does not request modifications to other skills or system-wide settings.

Assessment

This skill appears internally consistent and does not request credentials or install code. However: only submit email text you are comfortable sharing (avoid pasting passwords, sensitive personal data, or regulated health/financial information). If you follow the README's ClawHub or manual install instructions, verify the source before running any code. The SKILL.md links to an external 'context pack' site — review that site before providing any credentials or allowing the agent to fetch context automatically. If you plan to let an autonomous agent access your mailbox, ensure you understand and control how it obtains emails (this skill itself does not include mailbox access logic).

When asked to triage, sort, or process emails, follow this system.

Triage Categories

Assign every email exactly one category:

Category	Icon	Criteria
Urgent Action	🔴	Requires response/action within 24h. Deadlines, escalations, time-sensitive requests.
Action Required	🟡	Needs a response or task but not time-critical. Requests, approvals, questions.
FYI / Read	🔵	Informational. No action needed but worth reading. Updates, reports, announcements.
Delegate	🟣	Someone else should handle this. Forward with context.
Archive	⚪	Newsletters, automated notifications, receipts, spam-adjacent. No action needed.

Triage Output Format

For each email, produce:

[ICON] CATEGORY | From: sender | Subject: subject
Summary: One sentence — what this is and what's needed.
Action: Specific next step (or "None — archive")
Draft: [Yes/No] — whether a reply draft is included below

Draft Response Rules

Generate a reply draft when:

• Category is Urgent Action or Action Required
• The email contains a direct question or request
• User explicitly asks for drafts

Draft style:

• Be direct. Open with the answer or decision, not "Thank you for your email."
• Mirror their tone. Formal email gets formal reply. Casual gets casual.
• Keep it short. Most replies should be 2-5 sentences.
• End with clarity. What happens next? Who does what by when?
• Use the sender's name — never "Dear Sir/Madam" unless the original was that formal.

Batch Processing

When given multiple emails:

1. Triage all of them first — output the full sorted list grouped by category
2. Then provide drafts for Urgent Action and Action Required items
3. Highlight any patterns ("3 emails from the same client — might want a call instead")

Smart Signals

Flag these automatically:

• Repeated follow-ups from the same sender (they're waiting on you)
• CC escalation — when someone adds a manager or exec to the thread
• Deadline mentions — extract and highlight specific dates/times
• Sentiment shifts — if tone has gotten noticeably more terse or frustrated
• Meeting requests buried in email body (not calendar invites)

Action Item Extraction

Pull out discrete action items from emails:

• What needs to be done
• Who is expected to do it
• When it's due (if mentioned)
• Format as a checklist the user can copy into their task manager

When Not Triaging

If the user asks about a specific email (not batch triage), switch to focused mode:

• Summarize the email thread
• Identify the core ask
• Draft a response if requested
• Flag anything the user should know before replying

Pair with an industry context pack for domain-specific email handling (legal, healthcare, finance, etc.) at https://afrexai-cto.github.io/context-packs