monday

Cross-tool inbox aggregator. Runs each available source skill's monday sub-command in parallel, merges the JSON results, optionally rates each item with an AI agent, and prints a sorted JSON array on stdout. Progress messages go to stderr.

Source skills: gmail · slack · teams · outlook · github (as gh) · servicenow — each must be installed separately and expose its own [cmd] monday sub-command.

Quick start

# Auto-discover sources on PATH and aggregate the last 7 days, 50 items per source
monday

# Pick specific sources
monday gh slack --limit 20 --date 3d

# Aggregate + rate items (1-9 importance, 1-8 urgency, with summaries)
monday --rate-importance 9-1 --rate-urgency 8-1 --rate-summary 500

# Rate with a specific model and a knowledge-base context directory
monday gh --rate-importance 8-3 --rate-model claude-haiku-4-5 --rate-context /workspace/kb

Aggregation pipeline

The monday command executes the following steps in order:

1. Discover sources — with no positional args, run which [cmd] for each name in KNOWN_COMMANDS and keep those found on PATH; otherwise use the names supplied as positional args.
2. Invoke in parallel — call [cmd] monday --limit N --depth N --date Nd for every discovered source concurrently.
3. Collect and validate JSON — read each source's stdout; any source that exits non-zero, produces empty output, or emits invalid JSON is logged to stderr and dropped. Aggregation continues with remaining sources.
4. Deduplicate by id — merge all item arrays in source order; when two items share the same id, the first occurrence wins and later duplicates are dropped (so positional arg order determines precedence for overlapping sources).
5. Optionally rate — if any --rate-* flag is set, submit each item to the rating agent (model: --rate-model; optional context: --rate-context) to assign importance, urgency, and summary fields. Rating failures fall back to the unrated item; aggregation still completes.
6. Sort and output — if rated, sort by urgency × importance descending (ties broken by ts descending); otherwise sort by ts descending. Write the final JSON array to stdout.

Flags

Positional args (gh, slack, teams, outlook, gmail, servicenow) select which sources to invoke. With no positional args, monday runs which <cmd> on the known source list and uses whichever are found on PATH.

Source protocol

A command is "monday-compatible" when it accepts [cmd] monday --limit N --depth N --date Nd and writes a JSON array of items to stdout. Each item must include id (stable unique string for deduplication), ts (ISO timestamp for sorting), and any other fields (title, url, participants, body, etc.) which are passed through unchanged.

To add a new source, implement the protocol above and invoke it explicitly: monday [newcmd] gh slack. To register it for auto-discovery, add the command name to the KNOWN_COMMANDS list in the aggregator script.

Output

A single JSON array on stdout. Sorting and rating augmentation follow pipeline steps 5–6.

Unrated item example:

[
  {
    "id": "gh-pr-4821",
    "ts": "2025-07-14T09:12:00Z",
    "source": "gh",
    "title": "Fix race condition in auth middleware",
    "url": "https://github.com/org/repo/pull/4821",
    "participants": ["alice", "bob"]
  }
]

Rated item example (with --rate-importance 9-1 --rate-urgency 8-1 --rate-summary 200):

[
  {
    "id": "gh-pr-4821",
    "ts": "2025-07-14T09:12:00Z",
    "source": "gh",
    "title": "Fix race condition in auth middleware",
    "url": "https://github.com/org/repo/pull/4821",
    "participants": ["alice", "bob"],
    "importance": 8,
    "urgency": 7,
    "summary": "Security-critical PR awaiting your review; blocks the 2.4 release."
  }
]

monday gh slack --date 2d | jq '.[0:5] | .[] | {id, ts, title}'