Inbox Triage

Automates inbox management by categorizing messages into urgent, normal, or spam, generating daily digests, and drafting responses for low-priority items.

Security Scan

The skill's code is mostly a local triage/digest processor, but its description promises automated collection from multiple messaging platforms without requesting credentials or providing connectors — the capability claims do not match the actual implementation.

Purpose & Capability

The SKILL.md and README claim the skill will 'collect' messages from Signal, Telegram, Discord, email, etc., and integrate across platforms. The shipped code (scripts/triage.py) only processes messages passed via a JSON file and a local config; it contains no network connectors or calls and requires no API keys or tokens. Either the skill is incomplete (missing adapters) or its description overpromises platform integration.

Instruction Scope

Runtime instructions tell the agent to 'gather all messages from configured sources' and show platform examples, but the runtime code simply reads an input JSON and local config (~/.inbox-triage/config.yaml). The SKILL.md also instructs writing corrections to ~/.inbox-triage/corrections.log. There are no instructions that would exfiltrate data or call external endpoints, but the instructions are ambiguous and grant broad discretion about where to 'collect' messages without specifying how.

Install Mechanism

No install specification is provided (instruction-only packaging script present). There is no remote download or archive extraction, and included scripts are plain Python files. This minimizes install-time risk.

Credentials

The skill requests no environment variables or credentials. That is safer than asking for many secrets, but it's inconsistent with the claimed cross-platform integrations (which would normally require service tokens). The code does access user home paths (~/.inbox-triage) for config and corrections, which is proportionate for a local triage tool.

Persistence & Privilege

The skill persists configuration and a corrections log under the user's home directory (~/.inbox-triage/config.yaml and corrections.log). 'always' is false and the skill does not attempt to modify other skills or system-wide settings. Persisting user data in a hidden directory is expected for this type of tool but worth noting.

What to consider before installing

This package is a local triage/digest processor that expects input messages as a JSON file; it does not implement network connectors or request API keys for Signal/Telegram/email despite claiming cross-platform collection. Before installing/running: (1) Inspect scripts/triage.py yourself — it only reads an input JSON and writes to ~/.inbox-triage, and it does NOT auto-send replies (it only generates drafts). (2) If you expect automatic collection from messaging services, treat this skill as incomplete — it will not fetch messages without additional connectors and credentials. (3) Run the tests with synthetic data (test_messages.json) first, and verify outputs. (4) Be aware that it will create and write config/corrections files in your home directory; review those files and their contents if you care about privacy. If the missing platform adapters are intended, ask the publisher for the connector code or explicit instructions for providing messages securely; if you need automatic fetching from third-party services, prefer a skill that clearly documents required credentials and connection behavior.

Inbox Triage Skill

Automate message management by categorizing, summarizing, and drafting responses to keep your inbox clean and actionable.

Quick Start

# Install the skill
npx clawhub install inbox-triage

# Trigger
"Help me triage my inbox"

Core Features

1. Message Categorization

Messages are classified into three categories:

• Urgent: Requires immediate attention (deadlines, direct questions, emergencies)
• Normal: Important but can wait (updates, newsletters, routine items)
• Spam/Noise: Can be ignored or deleted (promotions, notifications, irrelevant)

Triggers for categorization:

• Contains time-sensitive language ("ASAP", "urgent", "deadline")
• Direct questions or requests for action
• From known contacts with time constraints

2. Daily Digest Generation

Creates a consolidated summary of all messages received that day:

📧 Daily Inbox Digest - April 15, 2026

🔴 URGENT (2):
  - Meeting reminder: Team sync at 3PM today
  - Question from Sarah: Need approval on budget by EOD

🟡 NORMAL (5):
  - Newsletter: Weekly tech roundup
  - Update: Project milestone reached
  - ...

🟢 SPAM/NOISE (12):
  - Promotions, notifications, alerts

3. Draft Response Generation

Auto-drafts replies for normal-spam categories:

• Acknowledgment: "Thanks for reaching out, I'll review and get back to you."
• Redirect: "This is outside my scope - try reaching out to X."
• Auto-reject: Polite decline for spam/unsolicited requests

When to Use This Skill

✅ Inbox/messages are overwhelming ✅ Need to sort through notifications and alerts ✅ Want to save time on routine responses ✅ Need daily summaries of important items ✅ Looking to filter spam automatically

❌ Not for high-stakes communication (legal, medical, financial advice) ❌ Not for creative work (writing, editing, brainstorming) ❌ Not for real-time conversations requiring immediate human response

How It Works

Input Processing

1. Collect: Gather all messages from configured sources
2. Analyze: Parse content, sender, timestamps, and context
3. Categorize: Apply priority rules and classification logic
4. Summarize: Generate digest or alert summaries
5. Draft: Create response options for review

Configuration

# Optional: ~/.inbox-triage/config.yaml
sources:
  - type: signal
    enabled: true
  - type: telegram
    enabled: true
  - type: discord
    enabled: false

rules:
  urgent_keywords:
    - "urgent"
    - "ASAP"
    - "deadline"
    - "important"
  spam_keywords:
    - "unsubscribe"
    - "promotion"
    - "offer"
  auto_draft_for:
    - "normal"
    - "spam"

Output Formats

JSON (machine-readable)

{
  "timestamp": "2026-04-15T12:00:00Z",
  "digest": {
    "urgent": [...],
    "normal": [...],
    "spam": [...]
  },
  "drafed_responses": [...]
}

Markdown (human-readable)

# Daily Triage Report

## 🔴 URGENT
- [ ] Item 1
- [ ] Item 2

## 🟡 NORMAL
- Item 1
- Item 2

## 🟢 SPAM
- 12 items filtered

Limitations

• Accuracy: Categorization is probabilistic, not perfect
• Context: May miss nuanced context in messages
• Human review required: Never auto-send without approval
• Platform support: Works best with text-based channels (Signal, Telegram, Discord, email)

Iteration

Track which categorizations were correct/incorrect:

# Log correction
echo "CORRECT: urgent - meeting reminder" >> ~/.inbox-triage/corrections.log
echo "INCORRECT: spam - actually important" >> ~/.inbox-triage/corrections.log

The system learns from corrections over time.

---

Integration Examples

With Cron Manager

# Run triage every morning at 8AM
0 8 * * * clawhub run inbox-triage --output daily-digest.md

With Weather Alert

# Send digest only when weather is clear
if [ "$(weather is-clear)" = "true" ]; then
  clawhub run inbox-triage --send-summary
fi

With File Organizer

# Attach digest as daily log
clawhub run inbox-triage --format json | tee ~/logs/daily-triage-$(date +%Y-%m-%d).json