slug: prompt-armor
provider: PromptArmor
generated_by: planning/capability-mapping/scripts/classify_capabilities.py
model: claude-opus-5
frame:
- Software & Technology
min_confidence: 0.7
capability_model:
  source: https://github.com/vincentmakes/turbo-ea-capabilities
  license: CC-BY-4.0
  attribution: Turbo EA Capabilities by Vincent Verdet — Turbo EA, https://github.com/vincentmakes/turbo-ea-capabilities,
    CC BY 4.0
  notice: NOTICE
edge_count: 2
edges:
- tag: Analyze
  spec_file: prompt-armor-analyze-api-openapi.yml
  capability_id: BC-620.30
  capability_id_l1: BC-620
  capability_name: Threat Detection & Response Management
  confidence: 0.72
  evidence: POST /v1/analyze/input "Analyze LLM input"; vendor "detects and blocks indirect prompt injection,
    data exfiltration, phishing, and system manipulation in production AI applications"
  reason: Real-time analysis of LLM traffic against threat detectors returning a block/allow verdict is
    threat detection and response, a cybersecurity capability. Ambiguity remains with AI governance framing,
    so not maximal confidence.
- tag: Content Check
  spec_file: prompt-armor-content-check-api-openapi.yml
  capability_id: BC-620.30
  capability_id_l1: BC-620
  capability_name: Threat Detection & Response Management
  confidence: 0.72
  evidence: POST /v1/check_content "Check content for adversarial / injected instructions"
  reason: Scanning content for adversarial/injected instructions is detection of attacks against the AI
    application — threat detection. Could alternatively be read as responsible-AI control, hence moderate
    confidence.