Skip to content

im-shashanks/claude-plugins / shaktra/skills/shaktra-quality

shaktra-quality

Verified

Quality engine for the TDD pipeline — provides quick-check (36 checks) for gate reviews and comprehensive review (14 dimensions) for final quality assessment. Loaded by sw-quality agent.

FileKindSize
SKILL.mdskill md3.5 KB
architecture-checks.mdmarkdown6.4 KB
comprehensive-review.mdmarkdown6.5 KB
performance-data-checks.mdmarkdown6.5 KB
quick-check.mdmarkdown13.1 KB
security-checks.mdmarkdown6.0 KB

comprehensive-review.md

View on GitHub
# Comprehensive Review

Full quality assessment for the QUALITY phase of the TDD pipeline. Uses 14 dimensions (A-M from `quality-dimensions.md` + Dimension N: Plan Adherence) plus coverage verification and decision consolidation.

**When:** QUALITY phase of TDD pipeline (Medium+ tiers). Thorough tier adds expanded analysis.

---

## Review Process

### 1. Load Context

Read before reviewing:
- Story YAML at `.shaktra/stories/<story_id>.yml`
- Handoff at `.shaktra/stories/<story_id>/handoff.yml`
- Implementation plan at `.shaktra/stories/<story_id>/implementation_plan.md`
- All files in `handoff.code_summary.files_modified`
- All files in `handoff.test_summary.test_files`
- `.shaktra/settings.yml` — for threshold values
- `.shaktra/memory/principles.yml` — for cross-story consistency

### 2. Run Tests and Verify Coverage

Execute the project's test suite using the configured `test_framework` and `coverage_tool` from settings.

**Verify:**
- All tests pass (`all_tests_green` confirmed by actual execution, not self-reported)
- Coverage meets or exceeds the tier threshold from `settings.tdd`
- Coverage comes from behavioral tests (not empty assertion tests that just execute code)

If tests fail or coverage is below threshold, emit `COVERAGE_GATE_FAILED` and stop the review — return to the developer for fixes.

### 3. Apply Dimensions A-M

Review using each of the 13 dimensions from `quality-dimensions.md`. For each dimension:

1. Apply the 3 key checks defined in the dimension
2. Check the P0 trigger condition
3. Record findings with: severity, dimension letter, file, line, issue description, guidance

**Dimension priority by scope:**
- Integration scope: Emphasize B (Failure Modes), F (Observability), G (Performance)
- Security scope: Emphasize A (Contract & API), E (Security), K (Configuration)
- Data scope: Emphasize C (Data Integrity), D (Concurrency), J (Deployment)

Skip dimensions that are clearly not applicable (e.g., Dimension D: Concurrency for a purely synchronous, single-threaded change). Document the skip reason.

### 4. Apply Dimension N — Plan Adherence

Verify the implementation matches the agreed plan:

**N.1 — Components Match Plan**
Compare `handoff.plan_summary.components` against actual files created. Every planned component must exist; every created file must trace to a planned component.
- Missing planned component → P0 (missing risk prevention)
- Extra unplanned component → P2 (document justification in code_summary)

**N.2 — Patterns Applied**
Verify each entry in `handoff.plan_summary.patterns_applied` was actually implemented.
- Pattern planned but not applied → P1
- Pattern applied differently → P2 (acceptable if justified in code_summary)

**N.3 — Scope Risks Mitigated**
For each risk in `handoff.plan_summary.scope_risks`, verify the prevention strategy was implemented.
- Risk identified but prevention missing → P0
- Prevention implemented differently → P2 (acceptable if justified)

**N.4 — Deviations Documented**
Any deviation from the plan must be documented in `handoff.code_summary` with justification.
- Undocumented deviation → P1

### 5. Quick-Check Cross-Reference

Read `handoff.quality_findings` — this contains findings from earlier quick-check gates (Plan, Test, Code), persisted by the orchestrator after each gate.

For each prior P0 or P1 finding where `resolved: false`:
- Verify the issue is resolved in the current code
- If still present: escalate to P0 (recurring finding after fix attempt)
- If resolved: confirm as addressed

If `quality_findings` is empty, skip this step (first-time review with no prior gates).

### 6. Observation Capture

Write observations about the review findings:

1. For each P0 or P1 finding: write a `quality-loop-finding` observation
2. For each fix applied: write a `fix-rationale` observation
3. For **every** principle and anti-pattern in `.briefing.yml`: write a `consistency-check` observation recording whether it was reinforced, weakened, or contradicted by this implementation. Include `principle_id` and `relationship`. This is mandatory — every briefing entry must have a corresponding consistency-check
4. Do NOT promote decisions to `decisions.yml` — the memory-curator handles all consolidation

### 7. Cross-Story Consistency

Compare this story's patterns against existing `principles.yml`:
- Does this implementation contradict a prior principle? → Flag as P1
- Does it establish a new pattern not yet recorded? → Recommend recording
- Does it deviate from an established pattern? → Flag as P2 if unjustified

---

## Thorough Tier — Expanded Review

For Large tier stories, add these additional review dimensions after the standard 14:

**Architecture Governance:**
- Apply ARC-01 through ARC-06 from `architecture-checks.md`
- For Large tier: additionally verify new modules don't fragment existing bounded contexts
- Cross-reference with `.shaktra/analysis/structure.yml` if analysis exists

**Performance Profiling Review:**
- Are there O(n^2) or worse algorithms on user-controlled input?
- Are database queries optimized (no N+1, appropriate indexing)?
- Are caching opportunities utilized where data is read-heavy?

**Dependency Audit:**
- Are all new dependencies actively maintained and appropriate for the use case?
- Do new dependencies introduce license incompatibilities?
- Could any new dependency be replaced with stdlib functionality?

---

## Output Format

```yaml
comprehensive_findings:
  - severity: P0|P1|P2|P3
    dimension: A|B|C|...|M|N
    check: "dimension check identifier"
    file: "path/to/file.py"
    line: 42
    issue: "Specific description of the problem"
    evidence: "Test result, log line, or code reference proving the issue"
    guidance: "Specific action to resolve"

coverage_result:
  tests_pass: true|false
  coverage_percent: integer
  tier_threshold: integer
  meets_threshold: true|false

observations_written:
  - type: "quality-loop-finding|fix-rationale|consistency-check"
    text: "observation summary"

cross_story_issues:
  - type: contradiction|new_pattern|deviation
    description: "what was found"
    existing_principle: "reference to principles.yml entry"
```

## Evidence Rule

Every finding must include evidence — a test result, log line, command output, or code reference. "This looks wrong" is never sufficient. "Line 42 calls `requests.get(url)` without a `timeout` parameter — see CQ-03" is.

## Gate Logic

After completing the review, apply the merge gate from `severity-taxonomy.md`:
- P0 > 0 → emit `QUALITY_BLOCKED`
- P1 > `settings.quality.p1_threshold` → emit `QUALITY_BLOCKED`
- Otherwise → emit `QUALITY_PASS`
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated 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