| name | add-source |
| description | Use when adding a new data source to the daily summary — MCP tools, CLI commands, local file paths, or HTTP endpoints. Triggered by "add source", "/add-source", "add to my daily summary", or "connect [tool] to my daily summary". |
Add Source
Add a new data source to your daily summary. The source will be checked automatically each time /daily-summary runs.
Step 0 — Locate custom-sources.md
Read memory/me.md for the workspace path. The file lives at {{WORKSPACE_PATH}}/memory/custom-sources.md. If it doesn't exist, create it from the template structure (see the format below).
Step 1 — Name and type
Ask: "What would you like to call this source? (e.g. 'Confluence team space', 'On-call incidents', 'Inbox folder')"
Ask: "What type of source is it?
- MCP tool — a connected tool like Confluence, PagerDuty, Intercom, etc.
- CLI command — a terminal command like
gh, jira, kubectl, a custom script
- Local files — a folder or file pattern on your computer
- HTTP endpoint — a URL you can query directly"
Step 2 — Access details
Based on type, ask:
MCP tool:
"What's the tool name and the call you'd use to query it? For example: confluence_search space=\"TEAM\" query=\"updated:today\". If you're not sure of the exact syntax, describe what you want and I'll help construct it."
CLI command:
"What's the command? I'll run it now to confirm it works:
{{COMMAND}}
Does the output look right?"
Run the command via Bash to verify it executes and returns something useful. If it errors, help the user fix it before proceeding.
Local files:
"What's the path or glob pattern? For example: ~/Documents/Inbox/*.md or /path/to/folder/. I'll check what's there now."
Glob the path to confirm files exist and show a sample.
HTTP endpoint:
"What's the URL and method (GET/POST)? Any headers or auth needed?"
Step 3 — What to extract
Ask: "What should the daily summary pull out from this source? Be specific — for example: 'pages updated today with title and author', 'open incidents with severity and assigned team', 'files modified today with a one-line summary of content'."
Step 4 — Filter
Ask: "How often should this be checked?
- Today only — only content from today
- Since last working day — today + yesterday (or Friday if today is Monday)
- Always — no date filter (good for things like open incident queues)"
Step 5 — Relevance
Ask: "Which projects or contexts is this relevant to? This helps the daily summary know where to include it.
- All projects
- Specific projects (list them)"
Step 6 — Preview and confirm
Show the complete entry that will be written:
## Source: {{NAME}}
- **Type:** {{TYPE}}
- **Access:** {{ACCESS}}
- **Extract:** {{EXTRACT}}
- **Filter:** {{FILTER}}
- **Relevant to:** {{RELEVANT_TO}}
Ask: "Does this look right? (yes / adjust)"
Step 7 — Write to custom-sources.md
Append the confirmed entry to memory/custom-sources.md above the closing comment line.
Step 8 — Test (optional)
Ask: "Want me to test this source now and show you what it would surface today? (yes / no)"
If yes, execute the source query and show sample output formatted as it would appear in the daily summary.
Step 9 — Confirm
"Added. '{{NAME}}' will be checked in every daily summary{{, scoped to: {{RELEVANT_TO}} if not 'all'}}. Run /daily-summary to see it in action."