summary

aweek:summary

Display a one-screen dashboard for every aweek agent. Each row shows the agent's name, goal count, weekly task progress, token-budget usage, and current lifecycle state.

This skill is read-only — it never mutates agent data.

Instructions

Follow this exact workflow when invoked. Use the project's Node.js modules in src/skills/summary.js — never read agent JSON files directly.

Step 1: Gather and render the dashboard

Run:

echo '{"dataDir":".aweek/agents"}' \
  | aweek exec summary buildSummary --input-json -

The JSON result exposes result.report (the formatted table) and result.agentCount. Present the formatted output from result.report directly to the user. Do not paraphrase or re-order columns — the table is intentionally compact.

Step 2: Offer drill-down (only when agents exist)

Do not volunteer the detail view automatically — keep the first screen focused on the dashboard table. When result.agentCount === 0, skip this step entirely.

2a. Fetch drill-down choices

Call getAgentDrillDownChoices to build a homogeneous list of options for the prompt (every real agent plus a synthetic No thanks entry with id: null):

echo '{"dataDir":".aweek/agents"}' \
  | aweek exec summary getAgentDrillDownChoices --input-json -

2b. Ask which agent to inspect

Use AskUserQuestion with the labels from the previous step. The recommended phrasing: "Inspect a specific agent in detail?"

• Map each choice's label to an answer option.
• The No thanks entry (with id: null) is always the last option — select it to end the skill without any detail view.

2c. Render the detail block

If the user picks a real agent, call buildAgentDrillDown with its id and print result.report verbatim:

echo '{"dataDir":".aweek/agents","agentId":"<id-from-the-user>"}' \
  | aweek exec summary buildAgentDrillDown --input-json -

The returned JSON has a report field — print it verbatim.

The detail report is generated by formatAgentStatus from src/skills/status.js — do not paraphrase or summarise it.

2d. Offer another drill-down (optional loop)

After showing the detail block, ask AskUserQuestion "Inspect another agent?" with the same choice list. Repeat step 2c until the user picks No thanks. Each loop iteration is a fresh AskUserQuestion — do not collect multiple picks in a single prompt.

Dashboard Columns

Column	Meaning
Agent	Agent display name (from identity.name)
Goals	Active goals (or active/total when some are completed/paused)
Tasks	Completed / total tasks in this week's plan (— when no plan)
Budget	used / limit (%) weekly tokens, or no limit when unlimited
Status	One of RUNNING, ACTIVE, PAUSED, IDLE

Status meanings

Value	Meaning
RUNNING	Agent has an active lock — a CLI session is executing now
ACTIVE	Agent has an approved plan with pending / in-progress tasks
PAUSED	Agent's weekly budget is exhausted — paused until reset or top-up
IDLE	Agent has no pending work or no approved plan

No Agents

If no agents exist, the output will suggest running /aweek:hire to create one. Do not attempt the drill-down step in that case.

Data Sources

All reads go through the file-based stores (files are the source of truth):

Source	Path Pattern
Agent config	.aweek/agents/<agent-id>.json
Weekly plans	.aweek/agents/<agent-id>/weekly-plans/<week>.json
Token usage	.aweek/agents/<agent-id>/usage/<monday>.json
Inbox queue	.aweek/agents/<agent-id>/inbox.json
Lock files	.aweek/.locks/<agent-id>.lock

Example Output

=== aweek Summary ===
Week: 2026-W16 (Monday: 2026-04-13)
Agents: 2

| Agent | Goals | Tasks | Budget                        | Status |
|-------|-------|-------|-------------------------------|--------|
| Alice | 2/3   | 2/5   | 25,000 / 100,000 (25%)        | ACTIVE |
| Bob   | 1     | 3/3   | 15,000 / 100,000 (15%)        | IDLE   |