Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Validates OSCAL System Security Plan documents against schemas, profiles, and cross-reference requirements with tiered validation depth.
.claude/skills/williamzujkowski-oscal-ssp-validator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 897% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 48% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 94% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 414% | 0% |
Trigger conditions:
Use this skill when:
Time normalization:
NOW_ET = <NIST time.gov semantics, America/New_York, ISO-8601>
Example: 2025-10-25T21:04:34-04:00Input validation:
ssp_path must resolve to accessible file or reachable URL.json, .xml, .yaml/.yml extension or MIME type)profile specified: must be valid OSCAL profile identifier or resolvable path/URLstrict must be boolean (default: false)Schema freshness check:
Goal: Confirm well-formedness and basic schema compliance
Steps:
metadata/version and metadata/oscal-versionvalid, invalid, or errorStop condition: If document is malformed or has >10 schema errors, abort and report.
Goal: Verify profile compliance and control implementation integrity
Steps (extends T1):
profile provided)control-implementation sectionOutput enhanced report:
Goal: Comprehensive integrity validation with detailed rationale
Steps (extends T2):
component-definition UUIDslink/@href referenceslast-modified vs. published datesstrict=false: warn on missing controlsstrict=true: fail on any control gapstrict=trueJSON Report Schema:
json{ "validation_timestamp": "ISO-8601 timestamp", "ssp_path": "string", "oscal_version": "string", "tier": "1|2|3", "status": "valid|invalid|error", "summary": { "total_errors": "integer", "total_warnings": "integer", "schema_valid": "boolean", "profile_aligned": "boolean|null", "cross_refs_valid": "boolean|null" }, "errors": [ { "type": "schema|profile|cross-ref|metadata", "severity": "error|warning", "location": "JSONPath or XPath", "message": "string", "suggestion": "string|null" } ], "profile_report": { "controls_required": "integer", "controls_implemented": "integer", "coverage_percent": "float", "missing_controls": ["string"] }, "sources": ["array of citation URLs with access dates"] }
Markdown Summary Format:
markdown# OSCAL SSP Validation Report **Document:** `{ssp_path}` **Timestamp:** {validation_timestamp} **OSCAL Version:** {oscal_version} **Status:** {status} ## Summary - Errors: {total_errors} - Warnings: {total_warnings} - Schema Valid: {schema_valid} - Profile Aligned: {profile_aligned} ## Details {top 5 errors with suggestions} ## Sources {clickable links with access dates}
Example 1: T1 validation of valid SSP
bash# Input ssp_path: "./my-ssp.json" profile: null strict: false # Execution 1. Parse my-ssp.json → success 2. Extract metadata/oscal-version → "1.1.2" 3. Fetch JSON schema v1.1.2 4. Validate → 0 errors # Output (JSON) { "validation_timestamp": "2025-10-25T21:04:34-04:00", "ssp_path": "./my-ssp.json", "oscal_version": "1.1.2", "tier": "1", "status": "valid", "summary": { "total_errors": 0, "total_warnings": 0, "schema_valid": true }, "errors": [] }
Token budgets (mandatory):
Safety:
Auditability:
Determinism:
Official NIST OSCAL Documentation:
Schemas:
Tools:
Profile Resolution Specification:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 5,391 | 3,656 | -32% | 1 | 1 | 0% | 344 | 3,428 | +897% | 0 | 0 | — |
case-02 | fail→pass | 15,312 | 11,457 | -25% | 1 | 1 | 0% | 3,541 | 5,244 | +48% | 0 | 0 | — |
case-03 | fail→pass | 15,893 | 8,567 | -46% | 1 | 1 | 0% | 3,807 | 4,748 | +25% | 0 | 0 | — |
case-04 | fail→fail | 10,472 | 13,612 | +30% | 1 | 1 | 0% | 2,300 | 5,766 | +151% | 0 | 0 | — |
case-05 | fail→fail | 13,198 | 18,609 | +41% | 1 | 1 | 0% | 3,072 | 5,857 | +91% | 0 | 0 | — |
case-06 | fail→fail | 6,454 | 11,278 | +75% | 1 | 1 | 0% | 771 | 5,156 | +569% | 0 | 0 | — |
case-11 | pass→pass | 15,594 | 2,813 | -82% | 1 | 1 | 0% | 1,255 | 3,076 | +145% | 0 | 0 | — |
case-07 | fail→pass | 9,806 | 5,195 | -47% | 1 | 1 | 0% | 1,802 | 3,498 | +94% | 0 | 0 | — |
case-08 | fail→pass | 13,116 | 5,801 | -56% | 1 | 1 | 0% | 737 | 3,787 | +414% | 0 | 0 | — |
case-09 | pass→pass | 8,569 | 6,531 | -24% | 1 | 1 | 0% | 1,590 | 4,012 | +152% | 0 | 0 | — |
case-10 | pass→pass | 5,778 | 3,670 | -36% | 1 | 1 | 0% | 1,135 | 3,230 | +185% | 0 | 0 | — |
case-12 | fail→pass | 23,926 | 2,923 | -88% | 1 | 1 | 0% | 1,681 | 3,125 | +86% | 0 | 0 | — |
case-13 | fail→pass | 25,371 | 3,341 | -87% | 1 | 1 | 0% | 2,230 | 3,222 | +44% | 0 | 0 | — |
case-14 | pass→pass | 8,289 | 3,523 | -57% | 1 | 1 | 0% | 1,667 | 3,147 | +89% | 0 | 0 | — |
case-15 | fail→pass | 5,787 | 3,287 | -43% | 1 | 1 | 0% | 1,181 | 3,277 | +177% | 0 | 0 | — |
case-16 | pass→pass | 11,647 | 7,771 | -33% | 1 | 1 | 0% | 2,197 | 4,211 | +92% | 0 | 0 | — |
case-17 | fail→pass | 11,600 | 7,088 | -39% | 1 | 1 | 0% | 2,085 | 3,888 | +86% | 0 | 0 | — |
case-18 | pass→pass | 12,272 | 8,528 | -31% | 1 | 1 | 0% | 2,449 | 4,398 | +80% | 0 | 0 | — |
case-19 | pass→pass | 3,339 | 2,346 | -30% | 1 | 1 | 0% | 623 | 2,993 | +380% | 0 | 0 | — |
case-20 | pass→pass | 12,810 | 4,407 | -66% | 1 | 1 | 0% | 2,098 | 3,289 | +57% | 0 | 0 | — |
case-21 | pass→pass | 10,386 | 3,942 | -62% | 1 | 1 | 0% | 1,691 | 3,152 | +86% | 0 | 0 | — |
case-22 | fail→fail | 8,710 | 2,561 | -71% | 1 | 1 | 0% | 1,683 | 2,989 | +78% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 18 counted toward the lift figure. The other 4 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +41 percentage points is the difference between those two pass rates over the 18 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.