This skill should be used when identifying, analyzing, and mitigating security risks in Artificial Intelligence systems using the CoSAI (Coalition for Secure AI) Risk Map framework. Use when assessing AI system security, conducting risk analysis for LLM applications, ML pipelines, model training/serving infrastructure, or generating compliance reports aligned with MITRE ATLAS, NIST AI RMF, OWASP Top 10 for LLM, and STRIDE frameworks. Supports both automated assessments and interactive exploration with 30+ query methods.
For requests like "Analyze security risks in [target]" or "Generate a CoSAI risk assessment":
uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/orchestrate_risk_assessment.py \
--target <user_specified_target> \
--output-dir ./risk-assessment-output
The orchestrator automatically:
For ad-hoc queries, threat modeling, or compliance mapping, use the CLI scripts. All support --offline flag for bundled schema usage.
| Purpose | Script |
|---------|--------|
| Search risks by keyword | uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/cli_risk_search.py "injection" |
| Search controls by keyword | uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/cli_control_search.py "training" |
| Get controls for a risk | uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/cli_controls_for_risk.py DP |
| Get persona risk profile | uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/cli_persona_profile.py personaModelProvider |
| Assess control coverage | uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/cli_gap_analysis.py DP --implemented controlTrainingDataSanitization |
| Get framework mappings | uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/cli_framework_map.py PIJ --framework mitre-atlas |
| Risks by actor access level | uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/cli_actor_access.py agent |
See references/exploration_guide.md for entity IDs, query patterns, and the core analyzer API.
Network/SSL Failures: Orchestrator automatically switches to bundled offline schemas — workflow continues without user intervention.
Manual offline mode:
uv run ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/orchestrate_risk_assessment.py --target <path> --offline
Missing Target: Ask user: "Please specify a file, directory path, or brief system description to analyze"
| Topic | Reference |
|-------|-----------|
| Framework fundamentals | references/cosai_overview.md |
| Persona definitions | references/personas_guide.md |
| Risk & schema structures | references/schemas_reference.md |
| Data collection forms | references/forms.md |
| Workflow procedures | references/workflow_guide.md |
| Interactive exploration | references/exploration_guide.md |
| Usage examples | references/usage_examples.md |
| Threat actor access levels | references/actor_access_guide.md |
Automation Scripts: (via ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/)
orchestrate_risk_assessment.py - Workflow orchestratorfetch_cosai_schemas.py - Schema downloaderanalyze_risks.py - Risk identification enginegenerate_report.py - Report generatorcore_analyzer.py - Core query API (30+ methods)Interactive CLI: (via ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/scripts/)
cli_risk_search.py - Search riskscli_control_search.py - Search controlscli_controls_for_risk.py - Controls for riskcli_persona_profile.py - Persona profilescli_gap_analysis.py - Gap analysiscli_framework_map.py - Framework mappingscli_actor_access.py - Risks by actor access levelBundled Assets: (via ${CLAUDE_PLUGIN_ROOT}/skills/ai-risk-mapper/)
assets/cosai-schemas/ - Offline schema cacheassets/report_template.md - Report templateExternal Framework: https://github.com/cosai-oasis/secure-ai-tooling
Edit PDFs with natural-language instructions using the nano-pdf CLI.
Control Sonos speakers (discover/status/play/volume/group).
Terminal Spotify playback/search via spogo (preferred) or spotify_player.
Capture frames or clips from RTSP/ONVIF cameras.
CLI to manage emails via IMAP/SMTP. Use `himalaya` to list, read, write, reply, forward, search, and organize emails from the terminal. Supports multiple accounts and message composition with MML (MIME Meta Language).
Monitor blogs and RSS/Atom feeds for updates using the blogwatcher CLI.
Category:tools