Digest of what your Superset agents did — sweeps workspaces, tasks, and agent terminals, then reports what finished, what needs review, and what's blocked.
SkillScore73Maintained
Automatic
◆Project Management pack
Last verified 8/29/2026
What it changes
Passed 12 of 17 test cases with the skill, and 7 of 17 without it.
Without skill41%
With skill73%
Checks it passed that the model alone missed
Sends input to the specified terminal
Does not conduct a multi-workspace standup sweep
Classifies the workspace under the Needs you category
and 4 more in the full report
8 fixed8 passed either way5 still failing1 broke
Pass-rate lift
+32 pp
Tokens
-47%
Agent turns
0%
Response time
-66%
Model gemini-3.6-flashMeasured 8/29/2026v116 of 22 cases passed with skill · 17 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 7 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 7
Prompt
An agent terminal in workspace ws-505 completed its task execution and is now waiting for code review. Under which summary bucket does this fall?
What was checked, without and with the skill
Checked
Without skill
With skill
Classifies the workspace under the Needs you category
Missed
Passed
case-20 · gemini-3.6-flash · 8/29/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md24 linessingle file
Superset Standup
Answer "what happened while I was away?" from real state, not guesses. Entirely read-only.
1. Sweep (parallel, prefer --json)
superset workspaces list — active workspaces
superset tasks list — task states
Per active workspace: superset terminals list --workspace <id>, then superset terminals read --workspace <id> --terminal <terminalId> for each agent terminal — the last screen of output shows whether the agent finished, asked a question, or errored
2. Classify each workspace
Needs you: agent finished and awaits review, asked a question, or hit a permission prompt / failure
In flight: actively working
Blocked: waiting on something external
Stale: idle with no pending work — candidate for cleanup
3. Report
Lead with what needs the user, one line per item: workspace, agent, state, and the next action. Then in-flight, then completed, then stale-workspace cleanup suggestions. Keep the whole digest scannable — no terminal dumps; quote at most the single relevant line an agent printed.
Rules
Never send input to a terminal, modify tasks, or clean anything up as part of the digest — offer those as follow-ups and act only when asked.