@governed/threats
v0.1.0
Published
Agentic AI threats with severity, taxonomy mappings, incidents, mitigations, and coverage gaps
Downloads
22
Maintainers
Readme
@governed/threats
Agentic AI threats with severity ratings, taxonomy mappings (OWASP, MITRE ATLAS, NIST), real-world incidents, mitigations, and framework coverage gaps.
Install
npm install @governed/threatsUsage
import {
threatsKnowledge,
getConfidenceStatus,
computeCoverageSummary,
getIncidentsForThreat,
getMitigationsForThreat,
getThreatsByCategory,
getAllGaps
} from '@governed/threats';
// Check data freshness
const status = getConfidenceStatus(threatsKnowledge);
console.log(status.status); // e.g. "Fresh - high confidence"
console.log(status.daysUntilStale); // days until re-evaluation needed
// Access all data directly
const { threats, incidents, mitigations, evolution } = threatsKnowledge;
// Filter threats by category
const memoryThreats = getThreatsByCategory(threatsKnowledge, 'memory-and-context');
// Get incidents linked to a specific threat
// IDs: TM-001 (memory poisoning), TM-002 (tool misuse), TM-003 (privilege escalation), TM-004 (multi-agent)
const relatedIncidents = getIncidentsForThreat(threatsKnowledge, 'TM-001');
// Get all mitigations (MIT-001, MIT-002, etc.), or those linked to a specific threat
const allMitigations = threatsKnowledge.mitigations;
const toolMisuseMitigations = getMitigationsForThreat(threatsKnowledge, 'TM-002');
// Coverage summary across frameworks
const coverage = computeCoverageSummary(threatsKnowledge);
// Find gaps in framework coverage
const gaps = getAllGaps(threatsKnowledge);What's included
- Threat catalog with severity ratings and rationale
- Cross-referenced taxonomy mappings (OWASP Agentic Top 10, MITRE ATLAS, NIST AI RMF)
- Real-world incidents with coverage analysis
- Mitigation references per threat
- Framework coverage gap analysis
- Confidence decay — data knows when it needs re-evaluation
Data freshness
Every export includes evaluation metadata: who evaluated, when, methodology, and validity window. Use getConfidenceStatus() to check programmatically.
Source
Registry source: registry/knowledge/threats.ts
Full component with visualization: living-governance.com
License: MIT
