dd-report

Report — Format & Deliver Analysis

Final skill in the pipeline. Reads the completed analysis object and formats it for the user's requested delivery channel.

Input: /tmp/dd-analysis/analysis.json (built by dd-analyse-core, optionally extended by dd-analyse-sap) Output: Formatted report delivered to the requested channel

---

Step 1 — Detect Delivery Intent

From the user's original message:

User says	Delivery
"send to Slack" / "DM me" / "post to #channel"	Slack via Slack MCP
"put it in Notion" / "create a page"	Notion page via Notion MCP
"write a report" / "save as doc"	.md or .docx file
"email the summary" / "send via email"	Outlook via Microsoft 365 MCP
(nothing specified)	Present in chat (default)

---

Step 2 — Load the Analysis Object

import json
with open('/tmp/dd-analysis/analysis.json') as f:
    analysis = json.load(f)

---

Step 3 — Format for Delivery Channel

Default Chat Format

**<CLIENT_CODE>** · <DD Mon> · <HH:MM – HH:MM AEST> (<duration>) · <Integration type> integration

**Health:** <total> logs — <error_rate> issues
Info <N> · Debug <N> · Warn <N> · Error <N>

**Fetch stats:** Phase 1 (error+warn): <N> · Phase 2 (info+debug): <N> · Pages: <N>

**What's happening:**
<2-4 sentence analyst interpretation>

**Top issues:**
1. <issue> — **<count>** hits
2. ...

**SAP / Integration:**
← INCLUDE THIS SECTION ONLY for SAP integrations where sap_integration exists.
← OMIT ENTIRELY for AMT, Maximo, generic, or when sap_integration is absent.
← Do NOT write "N/A" — simply leave the section out.
- <label> — <count> — WOs: ...

**Monitors:** (only if monitors is non-empty)
- <monitor name> — <status>

**Recent Events:** (only if events is non-empty)
- <event title> — <alert_type> — <date>

Important formatting rules:

• Only 4 log levels: error, warn, info, debug. Never include critical or emergency.
• SAP section: include ONLY for SAP integrations. Omit entirely for all others.
• Monitors/Events: include only if data was fetched and is non-empty.

Slack Format

Same structure adapted for Slack mrkdwn:

• Use *bold* instead of **bold**
• Use :warning: :red_circle: :white_check_mark: emojis for severity
• Keep concise and scannable
• Send via Slack:slack_send_message or Slack:slack_send_message_draft

*<CLIENT_CODE>* · <DD Mon> · <HH:MM – HH:MM AEST> · <Integration type>

*Health:* <total> logs — <error_rate> issues
Info <N> · Debug <N> · Warn <N> · Error <N>

*What's happening:*
<interpretation>

*Top issues:*
1. <issue> — *<count>* hits
2. ...

Notion Format

Create a Notion page with:

• Title: <CLIENT_CODE> Log Analysis — <DD Mon YYYY>
• Sections matching the chat format, using Notion rich text formatting
• Use Notion:notion-create-pages tool

Email Format (Outlook)

Subject: [Datadog] <CLIENT_CODE> Log Analysis — <DD Mon> Body: Same structure as chat format, formatted as HTML. Use Microsoft 365 MCP tools to send.

File Format (.md / .docx)

• .md: Write the chat format to a file
• .docx: Read the docx skill and produce a formatted Word document

---

Step 4 — Deliver

Execute delivery based on detected intent. For Slack and email, ask for user confirmation before sending.

For chat delivery, output the formatted text directly in the conversation.

---

Re-delivery

If the user asks to re-deliver a previous analysis ("send that to Slack", "now put it in Notion"), check if /tmp/dd-analysis/analysis.json still exists. If so, re-read it and format for the new channel. No need to re-run the pipeline.

If the file doesn't exist (new session), inform the user that the analysis needs to be re-run first.