| name | gza-summary |
| description | Synthesize operator triage guidance from failed-task history, unimplemented plan and explore work, unmerged branches, and queue state |
| allowed-tools | Bash(uv run gza history:*), Bash(uv run gza unmerged:*), Bash(uv run gza next:*), Bash(uv run gza advance:*), Bash(uv run gza watch:*) |
| version | 2.0.0 |
| public | true |
Gza Summary
Produce an operator triage summary by combining dedicated gza surfaces, then recommend the next concrete actions.
Process
Step 1: Gather task data
Run these commands to collect the canonical source data:
Failed-task history:
uv run gza history --status failed
Unimplemented plan/explore work:
uv run gza advance --unimplemented
Until uv run gza unimplemented exists, uv run gza advance --unimplemented is the canonical surface.
Unmerged code work:
uv run gza unmerged
Queue state, including blocked rows:
uv run gza next --all
Optional failed-recovery decision surface:
uv run gza watch --restart-failed --dry-run
Step 2: Analyze the output
Treat each command as the authoritative surface for one domain:
From uv run gza history --status failed:
- Treat this as factual failed-task history, including attempts that were later retried or resumed successfully
- Suggest factual follow-up like
uv run gza log <full prefixed task id> when operators need details
- Use
uv run gza watch --restart-failed --dry-run when operators need the current failed-task recovery decision surface
- Do not treat this as a replacement for the old mixed-bucket
gza incomplete output
From uv run gza unmerged:
- Each listed row is completed code work that still needs merge/review/sync attention
- Prefer gza-native guidance like
uv run gza merge <full prefixed task id> or uv run gza sync <full prefixed task id>
From uv run gza advance --unimplemented:
- Identify completed
plan or explore rows that do not yet have implementation work
- Suggest bare
uv run gza implement <full prefixed task id> only for completed plan rows; note that it queues by default
- Suggest
uv run gza advance --unimplemented --create when implement tasks should be queued from listed source rows
From uv run gza next --all:
- Identify the runnable next task(s) for
uv run gza work
- Call out blocked rows, including dropped-dependency blockers, as queue state rather than lineage state
From uv run gza watch --restart-failed --dry-run (optional):
- Use it only when failed tasks exist and the operator needs the failed-recovery decision surface
- Treat it as supplemental guidance, not as the primary summary source
Step 3: Suggest next steps
Produce a prioritized set of specific recommendations. Keep recommendations mapped back to the dedicated surface they came from.
Examples:
- "Inspect failed task
gza-1234 with uv run gza log gza-1234 before choosing a recovery action."
- "Queue implementation follow-up for
gza-1250 with uv run gza advance --unimplemented --create."
- "Merge unmerged implementation
gza-1301 with uv run gza merge gza-1301."
- "Run
uv run gza work to start the next runnable pending task from the queue."
Avoid broad restatements of task state. Focus on what the operator should do next and why.
Output format
Present the summary in these sections:
## Failed Recovery
## Unimplemented Plans/Explores
## Unmerged Work
## Queue State
## Suggested Next Steps
Keep it concise. If a section has no actionable items, say so briefly and move on.
Important notes
/gza-summary is the synthesized recommendation layer. It is not a canonical replacement for gza incomplete.
- Keep failed history, unmerged work, unimplemented follow-up, and queue state on their dedicated surfaces.
- Prefer gza-native commands in recommendations.
- Use full prefixed task IDs in every command example.