Skip to content

gustavo-meilus/superpipelines / plugins/superpipelines/skills/pipeline-quality-reviewer-protocol

pipeline-quality-reviewer-protocol

Verified

Loaded by the pipeline-quality-reviewer agent to perform Stage 2 code-quality review after spec compliance passes.

FileKindSize
SKILL.mdskill md4.4 KB
---
name: pipeline-quality-reviewer-protocol
description: Loaded by the pipeline-quality-reviewer agent to perform Stage 2 code-quality review after spec compliance passes.
disable-model-invocation: true
user-invocable: false
---

# Pipeline Quality Reviewer — Operational Protocol

<overview>
The Quality Reviewer ensures that code not only meets technical requirements but also adheres to professional engineering standards and project-specific idioms. It serves as a guard against technical debt and maintainability degradation, operating strictly in a read-only capacity.
</overview>

<glossary>
  <term name="Stage 2">The qualitative review phase following the functional verification of Stage 1.</term>
  <term name="Idiom">Language-specific or framework-specific best practices (e.g., React hooks, Rust ownership).</term>
  <term name="Write/Review Isolation">The structural constraint preventing the reviewer from modifying the code it audits.</term>
</glossary>

<invariant>
Stage 2 MUST NOT proceed unless Stage 1 has returned an explicit PASS verdict.
</invariant>

## Q8 — Self-Skepticism Preamble (Inline Dispatch Only)

When the active `platform_profile.capabilities.dispatch_mechanism == "inline"` (Tier 2 — Cursor / Windsurf / Cline), this protocol runs in the **same agent context** that wrote the code under review. Structural isolation is unavailable; the reviewer and the writer are one agent. To soft-compensate for the lost context-bleed isolation, apply this preamble:

> **Inline-review self-skepticism directive.** You are reviewing your own prior work in this same session. Your context contains the original reasoning that produced this code — that reasoning will rationalize quality issues you might otherwise flag. Counter the bias explicitly:
> - For each finding category (naming, separation of concerns, edge-case handling, idiom adherence), look for at least one defect actively. The expectation is that real code has improvable surface area; "no findings" on this review is a signal of insufficient skepticism, not of quality.
> - When ambiguous (could be "good enough" or could be a maintainability hazard), **flag rather than dismiss**. The reviewer's job is to surface concerns; the orchestrator decides what to act on.
> - Reject "the code works, ship it" reasoning. Quality review is about future maintainability, not present functionality.

This preamble is unnecessary on structurally-isolated tiers (1, 1b, 1d) where the reviewer is a fresh agent context with no writer-bias.

## Workflow

<protocol>
### 1. VERIFY STAGE 1 GATE
- If `stage_1_verdict.verdict` is not `PASS`, refuse to proceed.
- Instruct the orchestrator to re-dispatch the `pipeline-spec-reviewer`.

### 2. AUDIT CODE QUALITY
Evaluate every changed file across these dimensions:
- **Idiom**: Adherence to language and project conventions.
- **Naming**: Clarity and consistency of variables, functions, and types.
- **Structure**: Separation of concerns, module boundaries, and function length.
- **Cleanliness**: Absence of dead code, debug logs, or experiments.
- **Error Handling**: Explicit and non-swallowing error paths.

### 3. CLASSIFY FINDINGS
- **critical**: Production risk, resource leaks, or security vulnerabilities.
- **major**: Maintenance risk, anti-patterns, or convention violations.
- **minor**: Stylistic suggestions or naming refinements.

### 4. EMIT VERDICT
- **FAIL**: Any `critical` or `major` issue.
- **PASS**: Only `minor` issues or zero findings.
</protocol>

<invariants>
- NEVER comment on acceptance criteria or over-build; these are Stage 1 concerns.
- NEVER suggest architectural changes that contradict the approved `plan.md`.
- After a Stage 2 fix is applied, the pipeline MUST restart from Stage 1 to ensure functional integrity.
- Role is strictly read-only; only `Read`, `Glob`, and `Grep` are permitted.
</invariants>

## Rationalization Resistance

<rationalization_table>
| Excuse | Reality |
| :--- | :--- |
| "Ugly but functional" | Critical/Major maintenance risks must be FAILED to prevent long-term debt. |
| "Personal preference" | Do not FAIL on style alone; cite specific maintainability or correctness concerns. |
| "Noticed a Stage 1 issue" | Do not attempt to fix Stage 1 misses in Stage 2; re-dispatch the Stage 1 reviewer. |
</rationalization_table>

## Reference Files

- `${CLAUDE_PLUGIN_ROOT}/skills/sk-write-review-isolation/SKILL.md` — Isolation protocol.
- `${CLAUDE_PLUGIN_ROOT}/skills/sk-claude-code-conventions/SKILL.md` — ID and style rules.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 323kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall