Quality Gardening
Overview
Track quality as concrete metrics and debt movement, not as a vague single score.
This skill freezes new degradation first, then uses generated metrics to show current quality and convergence direction. For shared harness terms, see ../../references/harness-patterns.md; when quality surfaces are absent, use references/build-when-missing.md. For generated snapshot ownership, see references/generated-snapshot-ownership.md.
When To Use
- The user asks to design, evaluate, or maintain
QUALITY_SCORE.md, quality reports, quality snapshots, debt reduction, or red lines.
- A repository needs structural metrics beyond lint and tests.
- The team needs periodic gardening instead of a one-time large refactor.
Inputs Needed
- Current quality pain points: large files, scattered events, direct API calls, duplicated entrypoints, missing tests, or similar risks.
- Source paths to scan and paths to exclude.
- CI or scheduled run strategy.
Execution Order
- First: Find existing quality docs, generated reports, lint checks, and baselines.
- Then: Choose a small set of trackable metrics and thresholds; do not create a fake aggregate score.
- Finally: Output the quality report shape, gate policy, and convergence cadence.
Step-by-Step Process
- Search for
QUALITY_SCORE.md, quality scripts, baselines, and CI gardening workflows.
- Choose four to eight metrics, and make each one automatically collectable.
- If no quality report exists, bootstrap the minimum score/report/baseline set from
references/build-when-missing.md.
- Separate blocking gates from informational reports.
- Use thresholds or baselines for historical problems; new degradation should fail.
- Generate Markdown for humans and JSON for scripts and trends.
- Design a gardening cadence: lower one threshold or close one allowlist item at a time.
Checks
- Metrics: each metric is repeatable and reflects a real structural risk.
- Thresholds: checks freeze new degradation instead of demanding immediate full cleanup.
- Reports: each finding includes path, metric, current value, threshold, and suggested action.
- CI: it is clear which metrics fail builds and which only report.
- Incentives: do not hide critical risks behind one attractive total score.
Output Format
# Quality Gardening
## Detected Mapping
- quality:
- validation:
- work-state:
## Metrics
| Metric | Source | Threshold | Gate |
| --- | --- | --- | --- |
## Generated Reports
-
## Baseline Policy
-
## Garden Loop
-
## Non-Goals
-
Common Mistakes
- Inventing a good-looking total score that does not guide repairs.
- Tracking too many metrics, so nobody reads or fixes them.
- Skipping baselines, which makes the quality gate unusable on day one.
- Treating one large refactor as quality governance.
Example Prompts
- "Design QUALITY_SCORE.md and a quality report for this repo."
- "How should we track structural debt without inventing a fake score?"
- "Add historical large files and scattered calls to a quality garden."
1---2name: quality-gardening3description: Use when designing quality snapshots, generated quality reports, structural metrics, debt thresholds, regression budgets, quality gates, or gradual cleanup loops.4metadata:5 source: https://skillmd.com/skills/concertonotes/quality-gardening6---78# Quality Gardening910## Overview1112Track quality as concrete metrics and debt movement, not as a vague single score.1314This skill freezes new degradation first, then uses generated metrics to show current quality and convergence direction. For shared harness terms, see `../../references/harness-patterns.md`; when quality surfaces are absent, use `references/build-when-missing.md`. For generated snapshot ownership, see `references/generated-snapshot-ownership.md`.1516## When To Use1718- The user asks to design, evaluate, or maintain `QUALITY_SCORE.md`, quality reports, quality snapshots, debt reduction, or red lines.19- A repository needs structural metrics beyond lint and tests.20- The team needs periodic gardening instead of a one-time large refactor.2122## Inputs Needed2324- Current quality pain points: large files, scattered events, direct API calls, duplicated entrypoints, missing tests, or similar risks.25- Source paths to scan and paths to exclude.26- CI or scheduled run strategy.2728## Execution Order2930- First: Find existing quality docs, generated reports, lint checks, and baselines.31- Then: Choose a small set of trackable metrics and thresholds; do not create a fake aggregate score.32- Finally: Output the quality report shape, gate policy, and convergence cadence.3334## Step-by-Step Process35361. Search for `QUALITY_SCORE.md`, quality scripts, baselines, and CI gardening workflows.372. Choose four to eight metrics, and make each one automatically collectable.383. If no quality report exists, bootstrap the minimum score/report/baseline set from `references/build-when-missing.md`.394. Separate blocking gates from informational reports.405. Use thresholds or baselines for historical problems; new degradation should fail.416. Generate Markdown for humans and JSON for scripts and trends.427. Design a gardening cadence: lower one threshold or close one allowlist item at a time.4344## Checks4546- Metrics: each metric is repeatable and reflects a real structural risk.47- Thresholds: checks freeze new degradation instead of demanding immediate full cleanup.48- Reports: each finding includes path, metric, current value, threshold, and suggested action.49- CI: it is clear which metrics fail builds and which only report.50- Incentives: do not hide critical risks behind one attractive total score.5152## Output Format5354```markdown55# Quality Gardening5657## Detected Mapping58- quality:59- validation:60- work-state:6162## Metrics63| Metric | Source | Threshold | Gate |64| --- | --- | --- | --- |6566## Generated Reports67-6869## Baseline Policy70-7172## Garden Loop73-7475## Non-Goals76-77```7879## Common Mistakes8081- Inventing a good-looking total score that does not guide repairs.82- Tracking too many metrics, so nobody reads or fixes them.83- Skipping baselines, which makes the quality gate unusable on day one.84- Treating one large refactor as quality governance.8586## Example Prompts8788- "Design QUALITY_SCORE.md and a quality report for this repo."89- "How should we track structural debt without inventing a fake score?"90- "Add historical large files and scattered calls to a quality garden."