Claude Code Plugins

Community-maintained marketplace

Feedback

sop-dogfooding-quality-detection

@majiayu000/claude-skill-registry
4
0

SOP for detecting quality regressions during dogfooding runs and turning them into actionable fixes.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name sop-dogfooding-quality-detection
description SOP for detecting quality regressions during dogfooding runs and turning them into actionable fixes.
allowed-tools Read, Write, Edit, Bash, Glob, Grep, Task, TodoWrite
model sonnet
x-version 3.2.0
x-category quality
x-vcl-compliance v3.1.1
x-cognitive-frames HON, MOR, COM, CLS, EVD, ASP, SPC

STANDARD OPERATING PROCEDURE

Purpose

Identify quality regressions and latent issues while dogfooding, ensuring findings are evidenced, prioritized, and fed back into improvement loops.

Trigger Conditions

  • Positive: active dogfooding sessions, regression sweeps after releases, or monitoring new features for emergent issues.
  • Negative: isolated bug triage without self-application or pattern capture.

Guardrails

  • Confidence ceiling: Use Confidence: X.XX (ceiling: TYPE Y.YY) with ceilings {inference/report 0.70, research 0.85, observation/definition 0.95}.
  • Evidence-first: Record file:line, logs, metrics, or reproduction steps for each detected issue.
  • Structure-first: Update examples/tests to reflect newly detected regressions and their fixes.
  • Prioritization: Tag severity and blast radius; block release on critical regressions until resolved or waived with rationale.

Execution Phases

  1. Observation & Capture
    • Monitor outputs, logs, and behaviors during dogfooding; collect anomalies.
    • Normalize entries with severity, location, and reproduction notes.
  2. Validation & Classification
    • Reproduce findings; distinguish false positives and intentional behavior.
    • Map to categories (correctness, performance, UX, security, reliability).
  3. Remediation & Feedback
    • Propose fixes and owners; add tests to prevent recurrence.
    • Feed learnings into pattern retrieval and references.
  4. Confidence & Closure
    • Confirm fixes or document waivers; state residual risk and confidence with ceiling.

Output Format

  • Log of detected issues with evidence and severity.
  • Reproduction steps and validation results.
  • Remediation plan and test updates.
  • Confidence statement using ceiling syntax.

Validation Checklist

  • Evidence captured with location/steps for each issue.
  • False positives filtered; categories assigned.
  • Fixes/tests identified and owners named.
  • Patterns/references updated where applicable.
  • Confidence ceiling provided; English-only output.

Confidence: 0.70 (ceiling: inference 0.70) - SOP rewritten per Prompt Architect confidence discipline and Skill Forge structure-first detection loop.