Skip to content

vt-v-review

Content-level review of VMS vault files. Goes beyond schema validation to check factual accuracy of norm references, consistency of reifegrad assessments, appropriateness of VisiTrans-specific content, and German language quality. Produces per-file review findings: approve, revise, or escalate-to-human. When used with --cwp, acts as Step 4 and writes the review gate.

Plugin: vms
Category: Other
Command: /vt-v-review

VMS Content Review

This skill performs content-level review of VMS vault files — checking accuracy, consistency, and appropriateness beyond what automated schema validation covers.

Process

  1. Read references/review-checklist.md for domain-specific review criteria
  2. Read references/norm-accuracy-guide.md for common mistakes
  3. Read target files specified by the argument
  4. Review each file against the checklist:
  5. Factual accuracy of norm references
  6. Reifegrad assessment plausibility
  7. VisiTrans-specific content appropriateness
  8. Cross-reference correctness
  9. German language quality
  10. Classify each file: approve / revise / escalate-to-human
  11. For revise items: auto-fix via the appropriate domain agent, re-check, upgrade to approved
  12. For escalate items: present to user with specific question requiring judgment
  13. Report findings with specific issues and recommendations

Review Criteria

Norm Reference Accuracy

  • ISO 27001 control numbers match actual Annex A structure
  • VDA ISA references use correct chapter numbering
  • NIS2 article references are accurate
  • ISO 9001 chapter numbers match actual standard structure
  • DSGVO article references are correct

Reifegrad Plausibility

  • Does the claimed reifegrad match the described implementation?
  • Is a reifegrad of 3+ justified for a 15-person company?
  • Are there Nachweise to support the claimed level?
  • Is the Beschreibung der Umsetzung consistent with the reifegrad?

VisiTrans Specificity

  • Does the content reference actual VisiTrans systems and tools?
  • Are supplier references accurate (from the supplier directory)?
  • Is the scope appropriate for a SaaS company?
  • Are generic boilerplate statements replaced with specific details?

Cross-Reference Consistency

  • Do wiki-links point to the correct documents?
  • Are norm_refs consistent between related files?
  • Do TOM measures align with ISMS controls they reference?

Impact Analysis & Cascade Awareness

Run the impact analyzer on the file under review: 

python3 impact_analyzer.py "<vault_path>" "<target_file>"
Resolve the script path relative to TOOLKIT_ROOT.

For each referencing file: - Read its review_date from YAML frontmatter - Compare against the reviewed file's last git modification date (git log -1 --format=%ai -- "<target_file>") - If a referencing file's review_date is older than the reviewed file's last modification: flag as "stale review_date — may need consistency check"

Report: - "N Dateien referenzieren diese Datei. Davon haben M ein aelteres review_date als die letzte Aenderung." - List the M files with their review_date and the reviewed file's modification date - This is informational — include in the review findings as advisory

Language Quality

  • Professional German business language
  • No English mixed in (except technical terms)
  • Consistent terminology across files
  • No placeholder text or template markers remaining

Classification

Rating Meaning Action
approve Content is accurate, complete, and appropriate Ready for commit
revise Minor issues found, can be fixed automatically Agent can fix
escalate-to-human Requires human judgment (reifegrad, legal claims, VisiTrans specifics) Flag for user review

Output Format

## Review Results — YYYY-MM-DD

### Summary
| Rating | Count |
|--------|-------|
| Approve | N |
| Revise | N |
| Escalate | N |

### File Details

#### [filename.md] — APPROVE
No issues found.

#### [filename.md] — REVISE
- [ ] Issue: [description]
- [ ] Fix: [recommendation]

#### [filename.md] — ESCALATE
- Reason: [why human judgment needed]
- Question: [specific question for the user]

Open Brain Capture (Optional)

After completing the review, if the capture_thought MCP tool is available, capture the review outcome.

How: 1. Check if capture_thought tool is available. If not: skip silently. 2. Call capture_thought with: 

thought: "VMS review {date}: {N} files reviewed. Approved: {N}. Revise: {N}. Escalated: {N}. Key finding: {most significant issue}. Management system: {ISMS/DSMS/QMS}."
3. On timeout or error: log debug message and continue. Never fail the skill.

CWP Mode (--cwp CWP-NN) — Step 4 of 5

When invoked with --cwp, this skill acts as Step 4 of the numbered workflow.

  1. Display step banner: 

    ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 Step 4 of 5 — REVIEW             CWP-NN
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

  2. Prerequisite check: Read file_progress — warn if any files still have status: pending:

    "Warnung: N Dateien noch nicht entworfen. Step 3 zuerst abschliessen?" Allow override.

  3. Review all drafted files:

  4. Read file_progress, filter to status: drafted
  5. Review each file against the checklist (see above)

  6. Per file, update .content-state.yaml:

    • file_progress[file].status: reviewed
    • file_progress[file].reviewed_at: <timestamp>
    • file_progress[file].review_rating: approve | revise | escalate

  7. Auto-fix revise items:

  8. For each revise file: invoke the appropriate domain agent to fix
  9. Re-check the file
  10. If fixed: upgrade review_rating to approve, set status: approved
  11. If still issues: keep as revise (counts as unresolved)

  12. Partial fix outcome: If some revise items are fixed and some are not, the gate status is FAIL — any unresolved revise item blocks PASS. The revise count in the gate file records successfully auto-fixed items (not remaining failures). Remaining failures are listed under unresolved_revise in the gate file and must be addressed before re-running Step 4.

  13. Handle escalations:

  14. For each escalate file: present to user with the specific question

  15. After user responds: record escalation_reason and set acknowledged: true

  16. Run audit: Invoke /vt-v-audit --cwp CWP-NN for structural validation

  17. Write review gate: .cwp-review-gate.yaml in V004 repo root: 

    cwp: CWP-NN
status: PASS           # PASS if zero unresolved revise items; FAIL if any remain
date: "<timestamp>"
branch: feature/vms-content-cwp-NN
files_reviewed: <N>
ratings:
  approve: <count>
  revise: <count>       # successfully auto-fixed count (NOT remaining failures)
  escalate: <count>
unresolved_revise: []  # files still rated revise after auto-fix; non-empty → FAIL
escalations:
  - file: "<path>"
    reason: "<why>"
    acknowledged: true/false
audit_metrics:
  compliance_score: <N>
  content_depth_complete_pct: <N>
  cross_ref_rate_pct: <N>

  18. Update .content-state.yaml:

  19. Set phase: review
  20. Set review_gate: PASS | FAIL

  21. Update progress.files_reviewed count

  22. Display next step: 

    Review gate: PASS (N approved, M escalated)
Next → Step 5: /vt-v-cwp complete CWP-NN