Embed badge
Show
Style
[](https://versuz.dev/skills/yandy-r-claude-plugins-opencode-plugin-skills-quick-review)Free SKILL.md scraped from GitHub. Clone the repo or copy the file directly into your Claude Code skills directory.
npx versuz@latest install yandy-r-claude-plugins-opencode-plugin-skills-quick-reviewgit clone https://github.com/yandy-r/claude-plugins.gitcp claude-plugins/SKILL.MD ~/.claude/skills/yandy-r-claude-plugins-opencode-plugin-skills-quick-review/SKILL.md---
name: quick-review
description: Fast interactive review of uncommitted changes — prints findings inline,
writes nothing by default. "Apply fixes" hands findings to /quick-fix; "Save to
file" writes an artifact under docs/prps/reviews; "Write file and apply fixes" hands
off to /review-fix. For short changes where a full /code-review artifact is overkill.
Use when the user asks to "quick review", "fast review", "review what I have", "on-the-fly
review", or says "/quick-review".
---
# Quick Review
Interactive low-friction review of uncommitted changes. Output is **inline**.
The default fix path is artifact-free.
**Input**: `$ARGUMENTS`
**Core rule**: "Apply fixes" MUST NOT write a review artifact. It hands the
inline findings directly to `/quick-fix`. Only "Save to file" and
"Write file and apply fixes" write `docs/prps/reviews/quick-*.md`.
---
## Phase 0 — Parse flags
Extract flags from `$ARGUMENTS`:
| Flag | Effect |
| -------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `--parallel` | Fan out REVIEW across 3 standalone `code-reviewer` sub-agents (correctness, security, quality) and merge findings. Works in opencode, Cursor, and Codex bundles. |
| `--team` | (Claude Code only) Same 3-reviewer fan-out as a coordinated agent team with `spawn coordinated subagents`, shared `the todo tracker`, per-reviewer `track the task`, and coordinated shutdown. Heavier dispatch, richer communication. |
| `--yes` | Skip the confirmation prompt and behave as "Apply fixes" through `/quick-fix`. Mutually exclusive with `--save` and `--write-and-apply`. |
| `--save` | Skip the confirmation prompt and behave as "Save to file" (write artifact, print Next steps, exit). Mutually exclusive with `--yes` and `--write-and-apply`. |
| `--write-and-apply` | Skip the confirmation prompt, write the review artifact, then hand off to `/review-fix`. Mutually exclusive with `--yes` and `--save`. |
| `--severity <level>` | Forwarded to `/quick-fix` or `/review-fix` during hand-off. Valid: `CRITICAL`, `HIGH`, `MEDIUM`, `LOW`. Default: `HIGH`. Ignored on "Save to file" / "Discard". |
| `--no-worktree` | Accepted as a **no-op**. Quick mode never creates a worktree. Emit a note: `--no-worktree has no effect in quick mode.` |
Strip these from `$ARGUMENTS` and set `PARALLEL_MODE`, `AGENT_TEAM_MODE`,
`AUTO_YES`, `AUTO_SAVE`, `AUTO_WRITE_AND_APPLY`, and `MIN_SEVERITY` (default
`HIGH`). The remaining text MUST be empty; if not, abort with:
```
Error: /quick-review takes no positional argument (only flags).
```
**Validation**:
- `--parallel` and `--team` mutually exclusive -> abort with: `--parallel and --team are mutually exclusive. Pick one.`
- More than one of `--yes`, `--save`, `--write-and-apply` present -> abort with: `--yes, --save, and --write-and-apply are mutually exclusive. Pick one.`
- `--team` in a Cursor/Codex bundle (no `spawn coordinated subagents` tool) -> abort with: `--team is not supported in bundle invocations; use --parallel instead.`
- `--severity` value not one of `CRITICAL|HIGH|MEDIUM|LOW` -> abort with: `Invalid --severity value. Use CRITICAL, HIGH, MEDIUM, or LOW.`
---
## Phase 1 — GATHER
```bash
git diff --name-only HEAD
```
If the output is empty, print `Nothing to review.` and exit. Do NOT prompt,
do NOT write anything.
Capture the list of changed files as `CHANGED_FILES` for the REVIEW phase.
---
## Phase 2 — REVIEW
The shape of this phase depends on `PARALLEL_MODE` and `AGENT_TEAM_MODE`:
| Flags | Path |
| ----------------- | ------------------------------------------- |
| Neither set | **Path A** — single-pass review (default) |
| `PARALLEL_MODE` | **Path B** — 3 parallel sub-agent reviewers |
| `AGENT_TEAM_MODE` | **Path C** — 3-reviewer agent team |
Findings from all paths stay **in memory**. Do not write to disk here.
### Path A — Single-Pass Review (default)
Read each changed file in full and apply the **Local / Quick Review —
Single-Pass Checklist** and **Severity Rubric** from:
```
~/.config/opencode/shared/references/review-checklist.md
```
### Path B — Parallel Sub-Agent Review (`PARALLEL_MODE=true`)
Dispatch **3 standalone `code-reviewer` sub-agents in parallel** in a SINGLE
message with MULTIPLE `Agent` tool calls. Use the **Local / Quick Mode Roster**
and **Standard Findings Format** from:
```
~/.config/opencode/shared/references/review-checklist.md
```
Each reviewer prompt must include:
1. The list of changed files (`git diff --name-only HEAD`)
2. Its assigned focus and checklist items (from the roster table)
3. The severity rubric
4. A directive to return findings in the Standard Findings Format
Apply the **Merge Procedure** defined in the reference to produce a single
combined findings list for Phase 3.
### Path C — Agent Team Review (`AGENT_TEAM_MODE=true`, Claude Code only)
> **MANDATORY — AGENT TEAMS REQUIRED**
>
> You MUST follow the agent-team lifecycle. Every `Agent` call MUST include
> `team_name=` AND `name=`. See
> `~/.config/opencode/shared/references/agent-team-dispatch.md`.
Roster and category split identical to Path B.
#### C.1 Build the team name
Team name: `qrev-local-<YYYYMMDD-HHMMSS>`. Generate the timestamp once at the
start of Phase 2 and reuse it if the user later selects an artifact-writing
path.
#### C.2 Create the team
```
spawn coordinated subagents: team_name="qrev-local-<timestamp>", description="Quick review team for uncommitted local changes"
```
On failure, abort.
#### C.3 Register subtasks
Create 3 tasks in the shared task list (flat graph — reviewers are independent):
```
track the task: subject="correctness-reviewer: code-quality review of uncommitted changes", description="<full reviewer prompt>"
track the task: subject="security-reviewer: security review of uncommitted changes", description="<full reviewer prompt>"
track the task: subject="quality-reviewer: best-practices review of uncommitted changes", description="<full reviewer prompt>"
```
If any `track the task` fails -> `end the coordinated run`, abort.
#### C.4 Spawn the 3 reviewers
Single message, three `Agent` calls. Every call MUST include `team_name`, `name`
(matching the `track the task` subject prefix), `subagent_type="code-reviewer"`,
description, and a prompt equivalent to Path B's reviewer prompt plus a note
that the teammate shares a task list with two siblings (name them) and may
`send follow-up instructions` them on overlaps, and must call `update the todo tracker` to mark its task
complete before returning.
#### C.5 Monitor and collect results
Use `the todo tracker` until all 3 tasks are `completed`. Failure policy:
- All 3 error -> `end the coordinated run`, abort with a clear error.
- 1 or 2 error -> record "partial review — {role} did not complete" and proceed
with the remaining findings. Note the gap when printing inline.
#### C.6 Shutdown and cleanup
```
send follow-up instructions(to="correctness-reviewer", message={type:"shutdown_request"})
send follow-up instructions(to="security-reviewer", message={type:"shutdown_request"})
send follow-up instructions(to="quality-reviewer", message={type:"shutdown_request"})
end the coordinated run
```
Always `end the coordinated run` — even on abort or partial failure.
#### C.7 Merge findings
Apply the reference's Merge Procedure and pass the combined findings to Phase 3.
---
## Phase 3 — REPORT (inline)
Assign sequential finding IDs (`F001`, `F002`, ...) ordered by severity
(CRITICAL first) then by file path. Every finding gets `Status: Open`.
Compute:
- `critical_count`, `high_count`, `medium_count`, `low_count`
- `total_findings`
- `finding_file_count` (unique files with findings)
- `large_review = total_findings >= 5 || finding_file_count >= 3`
**Print** the review to stdout in the Review Artifact Format, **without**
writing a file. The printed block looks like:
```markdown
# Quick Review — Uncommitted Changes
**Reviewed**: <ISO date>
**Mode**: Quick (inline)
**Author**: local
**Branch**: <current-branch>
**Decision**: (pending user confirmation)
## Summary
<1-2 sentence overall assessment>
## Findings
### CRITICAL
- **[F001]** `file:line` — <description>
- **Status**: Open
- **Category**: <Security | Correctness | Code Quality | Best Practices>
- **Suggested fix**: <concrete fix>
### HIGH
- **[F002]** ...
- **Status**: Open
- ...
### MEDIUM
- **[F003]** ...
- **Status**: Open
- ...
### LOW
- **[F004]** ...
- **Status**: Open
- ...
## Files Reviewed
- `file1.ts` (Modified)
- `file2.ts` (Added)
```
**Omissions** (vs. the standard Review Artifact Format): no
`## Worktree Setup` section (quick mode never creates a worktree) and no
`## Validation Results` section (quick mode does not run toolchain checks).
After the artifact block, print:
```
Findings: [C] {critical_count} [H] {high_count} [M] {medium_count} [L] {low_count}
```
If `large_review=true`, also print:
```
Large quick review: {total_findings} findings across {finding_file_count} files.
You can still apply fixes directly, or write an artifact before applying fixes for an audit trail.
```
If the count is zero across all severities, skip Phase 4 entirely. Print
`No findings. Nothing to do.` and exit.
Keep the exact printed review block in memory as `INLINE_REVIEW_PAYLOAD` for
`/quick-fix` or artifact writing.
---
## Phase 4 — DECIDE (confirmation)
If `AUTO_YES=true` -> skip this phase, proceed directly to Phase 5 as if the
user selected "Apply fixes".
If `AUTO_SAVE=true` -> skip this phase, proceed directly to Phase 5 as if the
user selected "Save to file".
If `AUTO_WRITE_AND_APPLY=true` -> skip this phase, proceed directly to Phase 5
as if the user selected "Write file and apply fixes".
Otherwise, ask the user with `ask the user`.
For normal reviews (`large_review=false`):
```
question: "What should I do with this review?"
header: "Review action"
options:
- label: "Apply fixes"
description: "Hand findings directly to /quick-fix. No review file is written. (recommended)"
- label: "Save to file"
description: "Write docs/prps/reviews/quick-<timestamp>-review.md and stop."
- label: "Discard"
description: "Write nothing, exit"
```
For large reviews (`large_review=true`):
```
question: "This is a larger quick review. What should I do?"
header: "Review action"
options:
- label: "Apply fixes"
description: "Hand findings directly to /quick-fix. No review file is written."
- label: "Write file and apply fixes"
description: "Write docs/prps/reviews/quick-<timestamp>-review.md, then use /review-fix for an audit trail. (recommended)"
- label: "Save to file"
description: "Write docs/prps/reviews/quick-<timestamp>-review.md and stop."
- label: "Discard"
description: "Write nothing, exit"
```
Mark "Apply fixes" with `(Recommended)` when `large_review=false` and
`critical_count + high_count > 0`. Mark "Write file and apply fixes" with
`(Recommended)` when `large_review=true`.
**Text-mode fallback** (if `ask the user` is unavailable — Cursor/Codex
bundles without that tool): print the available choices explicitly and WAIT for
the user's next message. Parse their reply case-insensitively:
- `yes` / `apply` / `apply fixes` / `fix` -> "Apply fixes"
- `write and apply` / `write file and apply fixes` / `file and fix` / `artifact and fix` -> "Write file and apply fixes"
- `save` / `save to file` / `write` / `file` -> "Save to file"
- `no` / `discard` / `cancel` / `skip` -> "Discard"
- Anything else -> re-prompt once, then default to `Discard` if still ambiguous.
---
## Phase 5 — HAND-OFF
Branch on the user's selection (or auto-confirm flag).
### Selection: "Discard"
Print `Discarded. No artifact written.` and exit. Do NOT write any file.
### Selection: "Apply fixes"
Invoke the `quick-fix` skill inline via the `Skill` tool. Do NOT write an
artifact first.
- `skill`: `"quick-fix"`
- `args`: `"[--parallel] [--severity <level>] <INLINE_REVIEW_PAYLOAD>"`
Forward `--parallel` if `PARALLEL_MODE=true`. Forward `--severity <level>` when
`MIN_SEVERITY != HIGH` (HIGH is the default on quick-fix).
After the Skill call returns, print:
```
Fixes complete. No review artifact was written.
```
Exit.
### Selection: "Save to file"
Write the artifact to disk using the Phase 3 content plus the canonical header
block:
```bash
mkdir -p docs/prps/reviews
TIMESTAMP=$(date +%Y%m%d-%H%M%S) # reuse the Path C timestamp if set
REVIEW_FILE="docs/prps/reviews/quick-${TIMESTAMP}-review.md"
```
Write the same artifact content that was printed in Phase 3, but set
`**Decision**: COMMENT` in the header (since a review was produced but no
GitHub decision is being posted).
Print:
```
Quick review written to: docs/prps/reviews/quick-<TIMESTAMP>-review.md
Findings: [C] {critical_count} [H] {high_count} [M] {medium_count} [L] {low_count}
Next steps:
/review-fix docs/prps/reviews/quick-<TIMESTAMP>-review.md # apply fixes {recommended_single if 1-2 Open findings}
/review-fix docs/prps/reviews/quick-<TIMESTAMP>-review.md --parallel # fan out fixes {recommended_parallel if 3+ Open across 2+ files}
```
Annotate ONE command with `# <- recommended`:
- 1-2 Open findings -> recommend the single-pass form.
- 3+ Open findings spanning 2+ files -> recommend the `--parallel` form.
- Otherwise default to the single-pass form.
Exit.
### Selection: "Write file and apply fixes"
Step 1 — write the artifact to disk exactly as in "Save to file".
Step 2 — invoke the `review-fix` skill inline via the `Skill` tool:
- `skill`: `"review-fix"`
- `args`: `"<REVIEW_FILE> [--parallel] [--severity <level>]"`
Forward `--parallel` if `PARALLEL_MODE=true`. Forward `--severity <level>` when
`MIN_SEVERITY != HIGH`.
Step 3 — after the Skill call returns, print:
```
Fixes complete. Review artifact updated in place: docs/prps/reviews/quick-<TIMESTAMP>-review.md
```
Step 4 — exit.
**Note**: `--team` from quick-review is NOT forwarded to either fix workflow
automatically. The review-phase fan-out is about finding issues quickly; the
fix phase has its own execution decision.
---
## Important Notes
- **Apply fixes is artifact-free**: direct apply MUST route to `/quick-fix`
without writing `docs/prps/reviews/quick-*.md`.
- **Large reviews are advisory**: `5+` findings or `3+` finding files adds a
"Write file and apply fixes" choice. It does not force artifact creation.
- **Explicit artifact-backed apply**: only `--write-and-apply` or the matching
prompt choice writes the artifact and invokes `/review-fix`.
- **Idempotent filename**: the `quick-{YYYYMMDD-HHMMSS}-review.md` pattern
produces a new file per artifact-writing invocation — no overwrites.
- **Not a replacement for `/code-review`**: if you need toolchain
validation (typecheck/lint/test/build), a `## Validation Results` block,
worktree isolation, or a GitHub review posted, use `/code-review`
without `--quick`.
---
## Integration with ycc
- `/code-review --quick` delegates to this skill.
- `/quick-fix` consumes inline findings for artifact-free fixing.
- `/review-fix` consumes written artifacts only.
- `/prp-commit` or `/git-workflow --commit` can commit the fixes afterward.