All skills
simota avatar

/matrix

@c805268
by shingo imotasimota/agent-skills85 stars
15

Controlling combinatorial explosion across multi-dimensional axes: minimum coverage sets, execution plans, test/deploy/UX/risk prioritization. Use when scoping multi-axis combinations.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/matrix

This session only. Nothing lands on disk.

referencerisk-weighted-coverage.md

≈2.3k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Risk-Weighted Coverage Reference

Purpose: Prioritize combinations by Risk Priority Number (RPN = Severity × Occurrence × Detection) and Action Priority (AP) per AIAG-VDA 2019. Integrates FMEA output from omen to produce risk-sorted coverage sets with residual-risk reporting.

Scope Boundary

  • matrix risk-cover: RPN / Action Priority weighting over combinatorial coverage; FMEA integration (this document).
  • matrix prioritize (elsewhere): General Critical/High/Medium/Low prioritization without RPN integration.
  • matrix pairwise (elsewhere): Pure 2-way coverage without risk weighting.
  • omen (elsewhere): FMEA / pre-mortem analysis. risk-cover consumes omen's RPN output.
  • Sentinel / Breach (elsewhere): Security RPN specifically for attack surface.
  • rank (elsewhere): General ICE/RICE/WSJF prioritization for backlog items, not per-combination test selection.

Workflow

FMEA IN    →  pull RPN scores from omen if available
           →  otherwise derive RPN per axis/combination with team input

SCORE      →  RPN = Severity × Occurrence × Detection
           →  each factor 1-10 (AIAG 4th ed. scale)
           →  cap at RPN = 1000

CLASSIFY   →  Action Priority: H (high) / M (medium) / L (low)
           →  AIAG-VDA 2019 replaces RPN-only thresholds with AP matrix

WEIGHT     →  multiply combinatorial coverage priority by RPN
           →  RPN ≥ 200 or AP=H → mandatory inclusion
           →  RPN 100-199 or AP=M → include if under budget
           →  RPN < 100 or AP=L → include on residual budget

RESIDUAL   →  report uncovered combinations with their RPN
           →  residual-risk score = Σ (RPN × uncovered-flag)
           →  hand off residuals to omen for depth review

RPN Scoring

Severity (S) 1-10 scale

Score Description
10 Catastrophic: loss of life, major regulatory breach, total data loss
9 Very high: injury risk, legal penalty, production-wide outage
8 High: SEV1 incident, PII leak, top-tier SLA breach
7 Significant: SEV2, important-customer impact
6 Moderate: SEV3, noticeable degradation
5 Low-moderate: limited customer reports
4 Low: minor feature issue
3 Minor: cosmetic, workaround available
2 Very minor: internal only
1 None: no customer or internal impact

Occurrence (O) 1-10 scale

Score Description Probability
10 Almost certain > 1 in 2
9 Very high 1 in 3
8 High 1 in 8
7 Moderately high 1 in 20
6 Moderate 1 in 80
5 Low-moderate 1 in 400
4 Low 1 in 2,000
3 Very low 1 in 15,000
2 Remote 1 in 150,000
1 Nearly impossible < 1 in 1,500,000

Detection (D) 1-10 scale (inverse — 1 is best)

Score Description
10 No detection method exists
9 Very unlikely to detect before release
8 Remote chance of detection
7 Very low detection probability
6 Low (may be caught by smoke tests)
5 Moderate (regression tests likely to catch)
4 Moderately high (unit tests catch)
3 High (multiple test layers)
2 Very high (CI gates + production canaries)
1 Certain (100% detection before release)

RPN Calculation

RPN = S × O × D    (range: 1 - 1000)

Action Priority (AIAG-VDA 2019)

RPN alone can mislead (e.g., S=10, O=2, D=5 → RPN 100 looks moderate but S=10 is critical). AIAG-VDA 2019 replaces pure RPN thresholds with an Action Priority matrix.

AP Matrix (simplified)

Severity Occurrence Detection Action Priority
9-10 any any H
7-8 7-10 any H
7-8 4-6 7-10 H
7-8 4-6 1-6 M
7-8 2-3 any M
7-8 1 any L
4-6 7-10 7-10 H
4-6 7-10 1-6 M
4-6 4-6 any M
4-6 2-3 any L
4-6 1 any L
1-3 any any L

Use AP as the primary gate; use RPN as a tie-breaker within each AP bucket.

Risk-Weighted Coverage Selection

Step 1  Enumerate combinations (optionally from `pairwise` or `cover`).
Step 2  Assign S, O, D per combination (derive or pull from omen).
Step 3  Compute RPN and AP per combination.
Step 4  Sort by (AP desc, RPN desc).
Step 5  Selection rule:
        AP = H        → mandatory include
        AP = M        → include if under budget
        AP = L        → include on residual budget
Step 6  For pairwise 2-way covering subset, ensure every (P_i, P_j) pair
        is still covered at least once AFTER risk weighting.
        If risk weighting drops a pair, re-insert the cheapest covering case.
Step 7  Report residual risk: Σ RPN of uncovered combinations.

Residual Risk Reporting

Every selection inevitably leaves some combinations uncovered (by budget or pairwise reduction). Report what remains uncovered and why.

### Residual Risk Report

| Uncovered Combination | S | O | D | RPN | AP | Reason Uncovered |
|----------------------|---|---|---|-----|----|----|
| (browser=IE11, locale=JA, payment=wire) | 6 | 3 | 8 | 144 | M | pairwise reduction |
| (region=APAC, plan=free, feature=SSO) | 4 | 7 | 4 | 112 | M | budget |
| ... | ... | ... | ... | ... | ... | ... |

**Total residual RPN**: 412
**Residual AP counts**: H=0, M=5, L=12
**Recommendation**: 
  - H residual count must be 0 (mandatory rule). If > 0, request budget increase.
  - M residual count ≤ 5 (acceptable with sign-off).
  - L residual count: advisory only.

Hand off to omen for depth review on any AP=H uncovered or RPN ≥ 200 uncovered.

FMEA Integration with omen

When omen has already produced an FMEA:

  1. Ingest omen's FMEA table (failure mode, cause, S, O, D, RPN).
  2. Map each failure mode to combinatorial axes: "this failure is triggered by parameter combination {X, Y, Z}".
  3. Inherit RPN / AP for that combination.
  4. Any axis combination not covered in omen's FMEA defaults to team-estimated S/O/D (flag as "estimated, not FMEA-backed").

Output table must indicate source:

Source: [FMEA (omen-20250410-FMEA-03) | Estimated (team)]

Security-Specific RPN (hand off to Sentinel / Breach)

For attack-surface combinations, use CVSS-based scoring instead of generic S/O/D:

Factor Mapping
Severity CVSS Base Score (0-10)
Occurrence Attack Vector × Attack Complexity (simplified: 10 = network + low complexity; 1 = physical + high complexity)
Detection Log/WAF/IDS coverage (inverse; 1 = monitored, 10 = dark)

Output Template

## Risk-Weighted Coverage Plan

### Risk Inputs
- **Source**: [FMEA / Estimated / Hybrid]
- **omen reference**: [ID if applicable]
- **Team reviewers**: [names]

### Scoring Summary
| AP Bucket | Count | Total RPN |
|-----------|-------|-----------|
| H | [n] | [sum] |
| M | [n] | [sum] |
| L | [n] | [sum] |

### Selected Test Cases (AP desc, RPN desc)
| TC | Combination | S | O | D | RPN | AP | Source |
|----|-------------|---|---|---|-----|----|----|
| T-1 | (P1=a, P2=x, P3=1) | 9 | 5 | 7 | 315 | H | FMEA-03 |
| T-2 | ... | ... | ... | ... | ... | ... | ... |

### Pairwise Coverage After Weighting
- **Pairs covered**: [%]
- **Pairs dropped by weighting and re-inserted**: [count]
- **Final test count**: [N]

### Residual Risk
| Uncovered | S | O | D | RPN | AP | Reason |
|-----------|---|---|---|-----|----|----|
| ... | ... | ... | ... | ... | ... | ... |

- **Total residual RPN**: [sum]
- **H residual**: [count] (must be 0)
- **M residual**: [count]
- **L residual**: [count]

### Handoff
- **Next agents**: Radar (unit), Voyager (E2E), Siege (load), Sentinel/Probe (security), omen (depth review on residual)

Common Pitfalls

Pitfall Fix
Using RPN thresholds alone Use AP matrix (AIAG-VDA 2019)
Everything is "high RPN" Calibrate O and D using real incident data
No residual report Always report uncovered combinations + their RPN
Dropping H-AP pairs by mistake Re-insert covering case for every dropped pair
Single-person scoring RPN is team-estimated; single scores are noisy
Ignoring FMEA input Pull omen FMEA when available; do not re-estimate

Deliverable Contract

When risk-cover completes, emit:

  • Risk input source (FMEA reference or estimated).
  • S/O/D scores per combination with rationale.
  • RPN and AP per combination.
  • Selected test set (AP desc, RPN desc).
  • Pairwise coverage preservation check after weighting.
  • Residual risk report (uncovered combinations + their RPN / AP).
  • Budget check (H residual must be 0).
  • Handoff: Radar / Voyager / Siege / Sentinel / Probe / omen.

References

  • AIAG-VDA FMEA Handbook (2019) — Action Priority matrix
  • AIAG FMEA 4th Edition — RPN scoring scales (S/O/D 1-10)
  • NIST SP 800-30 — Risk Management Framework (risk likelihood × impact)
  • CVSS v3.1 / v4.0 — Common Vulnerability Scoring System (security RPN)
  • IEC 60812 — FMEA procedure
  • ISO 31000 — Risk management principles

Source: SKILL.md on GitHub

1 warning13d5 checks · Risk SAFE
  • Gen Agent Trust Hub13d

    The Matrix skill is a comprehensive tool for combinatorial testing design, providing robust frameworks for pairwise and high-strength interaction testing based on NIST and academic standards. It focuses on generating optimized execution plans and risk-weighted coverage sets without possessing any capabilities for code execution, network exfiltration, or unauthorized file access. No security risks were identified within the skill's instructions or reference materials.

  • Socket13d

    No alerts

  • Snyk13d

    Risk: LOW · No issues

  • Runlayer6mo

    1/11 files flagged

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at c805268. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated 2 weeks ago

README badge

README badge for simota/agent-skills/matrix