Slash Command

/queue-analyze

From fullstack-team

Complements the auto `quick analysis` that runs inside `/team-next`. Use this **before**

Popularity

Stars

Forks

Invocation

How this command is triggered — by the user, by Claude, or both

Slash command

/fullstack-team:queue-analyze

Model invocable

No pre-commands

Context Preview

The summary Claude sees in its command listing — used to decide when to auto-load this command

# Queue Analyze — Interactive Clarification

Complements the auto `quick analysis` that runs inside `/team-next`. Use this **before**
starting `/loop /team-next` to clear out `needs-human` tasks so the autonomous run has
as much `ready` work as possible.

**This command asks the user questions.** Do not call it from `/loop` or `/team-next`.

---

## Preconditions

1. `.work-state/queue.json` exists (run `/queue-sync` first if not).
2. `gh` authenticated.

---

## Arguments

- `--only <#id,#id>` — only analyze these issue IDs (otherwise all `needs-human` tasks).
- `--skip-repro` — don't atte...

Command Content

129 lines · ~1.1k tokens

Stats

LanguageShell

Stars3

Forks1

MaintenanceExcellent

Last CommitJun 20, 2026

Actions

View Source View Plugin View on GitHub View README

Stats

Actions

Queue Analyze — Interactive Clarification

Complements the auto quick analysis that runs inside /team-next. Use this before starting /loop /team-next to clear out needs-human tasks so the autonomous run has as much ready work as possible.

This command asks the user questions. Do not call it from /loop or /team-next.

Preconditions

.work-state/queue.json exists (run /queue-sync first if not).
gh authenticated.

Arguments

--only <#id,#id> — only analyze these issue IDs (otherwise all needs-human tasks).
--skip-repro — don't attempt bug reproduction via manual-qa; just ask questions.

Steps

1. Collect candidates

Read .work-state/queue.json. Select tasks where:

status == "needs-human", OR
status == "pending" and no ready flag (weren't touched by quick analysis yet)

If --only provided, filter by those IDs.

If set is empty → print "nothing to analyze" and exit.

2. Attempt reproduction (BUG_FIX tasks only, unless `--skip-repro`)

For each BUG_FIX in the set, launch manual-qa agent in parallel (one Agent call, all tasks batched):

For issue #<id>: <title>
Body: <body>
Try to reproduce on dev environment. Report:
- reproduced: yes/no
- steps used
- observed vs expected
- logs/screenshots relevant to root cause

Write each result to vibe-report/issue-<id>-reproduce.md.

If manual-qa reproduces the bug cleanly → the task may become ready without asking the user anything (mark ready: true in step 5).

If reproduction fails → the task stays in candidates, and step 3 will include a question like "couldn't reproduce with steps X — what am I missing?".

3. Build question batch

For each remaining candidate, generate at most 3 concrete questions based on its needs-human reason + body analysis. Examples:

Missing acceptance criteria → "How do we know #42 is done? [options: tests pass / manual check / metric X / Other]"
Scope ambiguity → "#51 — does this touch mobile too? [yes / no / Other]"
Failed repro → "#17 reproduce failed on steps X. What's different in your env? [Other text]"
Destructive action clarification → "#89 requires dropping column Y. Confirm with backup plan? [confirmed / let's migrate instead / needs-human forever]"

Skip questions that are already answered in the issue body. Don't re-ask the obvious.

Flatten all questions across all issues into a single list, each tagged with its issue ID.

4. Ask the human (single batch)

Use AskUserQuestion with the whole batch at once — grouped visually by issue:

[#42] Acceptance criteria?      → [tests / manual / metric / Other]
[#42] Breaking API change OK?   → [yes / no / Other]
[#51] Scope includes mobile?    → [yes / no / Other]
[#17] Repro env details?        → [Other text]

One round, one interaction. If the user aborts / doesn't answer → leave tasks as they are (no status change) and exit.

5. Apply answers

For each task:

Write/update vibe-report/issue-<id>-context.md with the answers + any repro output.
If all questions answered AND no destructive/security blocker remains → ready: true, status: "pending" (so /team-next picks it up).
If user chose "needs-human forever" or skipped → leave as needs-human.
If a destructive/security blocker remains → keep needs-human with updated reason.

Commit queue + context files: chore(queue): human-analysis <date>.

6. Summary

Print a compact table:

Analyzed: 7 tasks
  → ready:        5  (#42, #51, #63, #64, #72)
  → needs-human:  2  (#17 — repro still unclear, #89 — awaiting DBA signoff)

Ready to run: /loop /team-next

Guardrails

Maximum 3 questions per issue. If a task genuinely needs more → mark it needs-human with reason "too complex for batch analyze, use /team interactively".
Never auto-resolve destructive/security tasks without an explicit "confirmed" answer.
Do not modify tasks outside the analyzed set.
Do not call /team or /team-next from here.

Start analysis now.

/queue-analyze

Popularity

Invocation

Context Preview

Command Content

/queue-analyze

Popularity

Invocation

Context Preview

Command Content

Queue Analyze — Interactive Clarification

Preconditions

Arguments

Steps

1. Collect candidates

2. Attempt reproduction (BUG_FIX tasks only, unless --skip-repro)

3. Build question batch

4. Ask the human (single batch)

5. Apply answers

6. Summary

Guardrails

Queue Analyze — Interactive Clarification

Preconditions

Arguments

Steps

1. Collect candidates

2. Attempt reproduction (BUG_FIX tasks only, unless --skip-repro)

3. Build question batch

4. Ask the human (single batch)

5. Apply answers

6. Summary

Guardrails

2. Attempt reproduction (BUG_FIX tasks only, unless `--skip-repro`)

2. Attempt reproduction (BUG_FIX tasks only, unless `--skip-repro`)