UnitOneAI · DENGXUELIN · Jun 8, 2026
diff --git a/skills/compliance/nist-csf-assessment/SKILL.md b/skills/compliance/nist-csf-assessment/SKILL.md
@@ -348,6 +348,51 @@ Determine the overall organizational Tier based on aggregated assessment across
 
 ---
 
+### Step 4.5: Profile Evidence Confidence Gate
+
+**Objective:** Score current and target profiles using evidence quality, scope,
+freshness, and risk context. Do not let interview-only statements, stale
+artifacts, or target-state aspirations carry the same confidence as current
+implementation evidence.
+
+Apply these gates before finalizing subcategory scores:
+
+```
+CSF-CONF-01: Subcategory score lacks evidence source, artifact ID, owner, evidence date, and assessed scope
+CSF-CONF-02: Evidence type is not classified (automated telemetry, independent test, audit artifact, owner attestation, interview, or plan)
+CSF-CONF-03: Current profile evidence is mixed with target profile intent or planned improvements
+CSF-CONF-04: Evidence freshness is unknown or outside the subcategory's acceptable review window
+CSF-CONF-05: Evidence coverage is incomplete for the business unit, system, supplier, data class, or geography being scored
+CSF-CONF-06: Evidence confidence ignores inherent risk, control criticality, or compensating evidence
+CSF-CONF-07: Assumptions and validation-needed items are omitted from low or medium confidence scores
+CSF-CONF-08: Remediation priority is based only on score gap and does not consider confidence or evidence weakness
+```
+
+**Evidence confidence model:**
+
+| Evidence Type | Typical Confidence | Notes |
+|---|---|---|
+| Automated telemetry / continuous control | High when current, scoped, and mapped to the subcategory |
+| Independent test or audit artifact | High/Medium depending on recency, scope, and method |
+| Owner attestation with manager sign-off | Medium by default; can be High for low-risk scope with corroborating artifacts |
+| Interview or workshop statement | Low until backed by artifacts or testing |
+| Target-state plan or roadmap | Useful for target profile only; not current implementation evidence |
+
+**Minimum profile evidence fields:**
+
+| Field | Required Evidence |
+|---|---|
+| Subcategory | Official CSF 2.0 ID and description |
+| Profile side | Current profile, target profile, or gap-analysis assumption |
+| Source | Artifact/report/tool/export/interview, owner, evidence date, and collection method |
+| Scope | Business unit, system, supplier, data class, geography, and exclusions |
+| Evidence type | Telemetry, independent test, audit artifact, owner attestation, interview, plan |
+| Score rationale | Why the evidence supports the current score and target score separately |
+| Risk context | Inherent risk, criticality, compensating evidence, and dependency strength |
+| Confidence | High/medium/low with assumptions and validation needed |
+
+---
+
 ### Step 5: Organizational Profile Development
 
 #### 5.1 Current Profile
@@ -358,6 +403,12 @@ Document the current state for each function/category/subcategory:
 | Function | Category | Subcategory | Current Score | Evidence | Gaps |
 ```
 
+For each current-profile score, add evidence metadata:
+
+```
+| Subcategory | Current Score | Evidence Type | Source / Owner / Date | Scope / Coverage | Confidence | Assumptions / Validation Needed |
+```
+
 #### 5.2 Target Profile
 
 Define the target state based on:
@@ -370,6 +421,12 @@ Define the target state based on:
 | Function | Category | Subcategory | Current Score | Target Score | Gap | Priority |
 ```
 
+Keep target-profile intent separate from current evidence:
+
+```
+| Subcategory | Target Score | Target Driver | Current Evidence Confidence | Planned Evidence Needed | Priority Impact |
+```
+
 #### 5.3 Gap Analysis
 
 For each subcategory where Current < Target:
@@ -459,6 +516,12 @@ Use the NIST CSF 2.0 Reference Tool for comprehensive mappings.
 | GV.OC-01 | Organizational mission informs CSRM | [0-4] | [0-4] | [delta] | [H/M/L] | [refs] |
 | ... | ... | ... | ... | ... | ... | ... |
 
+## Profile Evidence Confidence
+
+| Subcategory | Profile Side | Evidence Type | Source / Owner / Date | Scope / Coverage | Score Rationale | Risk Context | Confidence / Validation Needed |
+|-------------|--------------|---------------|-----------------------|------------------|-----------------|--------------|--------------------------------|
+| [CSF ID] | [current/target] | [type] | [artifact + owner + date] | [scope] | [why score is supported] | [risk/criticality] | [H/M/L + next validation] |
+
 ### IDENTIFY (ID)
 [same table format]
 
@@ -576,6 +639,8 @@ Tier 4 — Adaptive
 
 4. **Failing to develop actionable organizational profiles.** The current and target profiles are the primary outputs of a CSF assessment. Many organizations conduct the assessment but do not formalize profiles into living documents that drive investment decisions, resource allocation, and progress tracking. Without profiles, the assessment becomes a one-time exercise rather than a continuous improvement tool.
 
+5. **Treating unverified statements as control evidence.** Interviews and self-attestations can support a CSF assessment, but they need owner/date/scope, confidence, assumptions, and validation-needed metadata. Target-state roadmaps should not raise current-profile scores.
+
 ---
 
 ## Prompt Injection Safety Notice

diff --git a/...nce/nist-csf-assessment/tests/benign/risk-scoped-profile-confidence-evidence.md b/...nce/nist-csf-assessment/tests/benign/risk-scoped-profile-confidence-evidence.md
@@ -0,0 +1,46 @@
+# Benign Fixture: Risk-Scoped Profile Confidence Evidence
+
+## Scenario
+
+The assessment scores `PR.PS-01` for a non-critical internal reporting service.
+The evidence is a signed owner attestation backed by a configuration export and
+recent vulnerability-scan sample. The target profile remains higher, but the
+current score and confidence are documented separately.
+
+## Evidence Snapshot
+
+| Field | Value |
+|---|---|
+| Subcategory | `PR.PS-01` managed asset configuration is established and maintained |
+| Profile side | Current profile |
+| Current score | `2` |
+| Target score | `3` |
+| Evidence type | Owner attestation with corroborating artifacts |
+| Source artifact | `cfg-export-reporting-2026-06-20.json` |
+| Attestation | `control-owner-signoff-PRPS01-2026-Q2.pdf` |
+| Scan sample | `vm-scan-reporting-2026-06-22.csv` |
+| Owner | Reporting Platform Manager |
+| Evidence date | `2026-06-22` |
+| Scope | Internal reporting service, low inherent risk, no regulated data |
+| Coverage | 11 of 11 reporting hosts and IaC baseline |
+| Score rationale | Configuration baseline exists and is approved, but enforcement is not fully automated |
+| Risk context | Low criticality lowers required evidence threshold for current score `2` |
+| Confidence | Medium; acceptable for low-risk current score, validation needed for target `3` |
+| Validation needed | Prove automated drift detection before target score `3` |
+
+## Positive Controls
+
+- `CSF-CONF-01`: Evidence source, owner, date, and assessed scope are recorded.
+- `CSF-CONF-02`: Evidence type is classified and supported by artifacts.
+- `CSF-CONF-03`: Current score `2` is separate from target score `3`.
+- `CSF-CONF-04`: Evidence freshness is current for the quarterly assessment.
+- `CSF-CONF-05`: Coverage is explicit for all reporting hosts in scope.
+- `CSF-CONF-06`: Confidence considers low inherent risk and corroborating evidence.
+- `CSF-CONF-07`: Validation needed for target maturity is documented.
+- `CSF-CONF-08`: Remediation priority can account for both score gap and medium confidence.
+
+## Expected Result
+
+Accept current score `2` with medium confidence for the low-risk scope. Do not
+flag the owner attestation as invalid by itself, because it is scoped, dated,
+signed, corroborated, and paired with validation needed for the higher target.
diff --git a/...t-csf-assessment/tests/vulnerable/target-profile-treated-as-current-evidence.md b/...t-csf-assessment/tests/vulnerable/target-profile-treated-as-current-evidence.md
@@ -0,0 +1,52 @@
+# Vulnerable Fixture: Target Profile Treated as Current Evidence
+
+## Scenario
+
+The assessment scores `DE.CM-01` at current score `3` because leadership says the
+organization is targeting Tier 3 monitoring this year. The only artifact is a
+roadmap slide. There is no current telemetry, monitoring coverage inventory, or
+validation that monitored services match the assessment scope.
+
+## Evidence Snapshot
+
+| Field | Value |
+|---|---|
+| Subcategory | `DE.CM-01` networks and network services are monitored |
+| Profile side | Current profile |
+| Current score assigned | `3` |
+| Target score | `3` |
+| Evidence type | Target-state plan |
+| Source | `2026-security-roadmap.pptx` |
+| Owner | VP Security |
+| Evidence date | `2026-01-10` |
+| Assessment date | `2026-06-30` |
+| Scope claimed | Enterprise network monitoring |
+| Coverage proof | Not provided |
+| Telemetry proof | Not provided |
+| Assumptions | Not recorded |
+| Confidence | Not recorded |
+
+## Problem Indicators
+
+- `CSF-CONF-01`: The score lacks an implementation artifact, collection method,
+  and scoped evidence supporting the current state.
+- `CSF-CONF-02`: A target-state plan is not classified separately from current
+  implementation evidence.
+- `CSF-CONF-03`: Current profile score is inflated by target profile intent.
+- `CSF-CONF-04`: Evidence freshness is weak for a monitoring capability.
+- `CSF-CONF-05`: Enterprise coverage is claimed without asset/service scope.
+- `CSF-CONF-07`: Assumptions and validation-needed items are omitted.
+- `CSF-CONF-08`: Priority is understated because the low-confidence evidence is
+  not included in remediation ranking.
+
+## Expected Finding
+
+Classify as **Significant Gap** or mark current score as low-confidence until
+fresh monitoring telemetry, service coverage, owner attestation, and validation
+evidence are collected. The target score can remain `3`, but the current score
+should not be raised by roadmap intent.
+
+## Required Remediation
+
+Separate current and target evidence. Add source owner/date/scope, evidence type,
+coverage, confidence, assumptions, and validation-needed fields for `DE.CM-01`.