crucible

pip install crucible-security

crucible init --target https://my-agent.com/api/chat
crucible scan --target https://my-agent.com/api/chat
crucible report crucible-report.json

python examples/test_langchain_agent.py

python examples/test_openai_assistant.py

# Generate config
crucible init --target URL --provider openai --key sk-xxx

# Run a standard scan
crucible scan \
  --target https://my-agent.com/api/chat \
  --name "My ChatBot" \
  --header "Authorization: Bearer sk-xxx" \
  --timeout 30 \
  --concurrency 5

# Run with payload mutation (bypass WAFs/guardrails)
crucible scan --target URL --mutate

# Multi-turn attack strategy
crucible scan --target URL --strategy multi-turn

# Use agent profile to target attacks
crucible profile --target URL --output agent_profile.json
crucible scan --target URL --profile agent_profile.json

# Behavioral integrity audit (multi-turn drift detection)
crucible behavioral-audit \
  --target https://my-agent.com/api/chat \
  --baseline-turns 5 \
  --probe-turns 15

# Generate EU AI Act compliance report from scan results
crucible scan --target URL --output json > results.json
crucible compliance-report --results results.json --output compliance.md

# JSON output for CI/CD
crucible scan --target URL --output json > report.json

# Local model scanning (Ollama, LM Studio, HuggingFace TGI)
crucible scan --target http://localhost:11434 --format-preset ollama --model llama3

# Global rate limiting (2 requests per second)
crucible scan --target URL --rate-limit 2

# Scope enforcement via YAML file
crucible scan --target URL --scope-file scope.yaml

# Audit an MCP server for tool poisoning, command injection & OAuth scope abuse
crucible mcp-scan --server https://my-mcp.example.com

# With auth header and JSON output
crucible mcp-scan --server http://localhost:3000 \
  --header "Authorization: Bearer sk-xxx" \
  --output mcp-report.json

# Re-render a saved report
crucible report report.json

# Run scan with bootstrap statistical confidence intervals (calculate 95% CI with 10 runs per attack)
crucible scan --target URL --confidence --confidence-runs 10

# Validate a trace policy YAML file
crucible trace validate-policy policy.yaml

# Start the MCP interception & auditing trace proxy (plain HTTP)
crucible trace start --listen 8080 --upstream http://localhost:8001 --policy policy.yaml --log audit.jsonl

# Start the proxy with native TLS termination (auto-generated self-signed dev certificate)
crucible trace start --listen 9443 --upstream http://localhost:8001 --policy policy.yaml --tls-self-signed

# Start the proxy with native TLS termination (using custom certificate/key files)
crucible trace start --listen 9443 --upstream http://localhost:8001 --policy policy.yaml --tls --tls-cert cert.pem --tls-key key.pem

# Render a summary report from a trace audit log file
crucible trace report audit.jsonl

# Plant a poisoned document using Semantic Anchor injection (Technique 1)
crucible poison-test plant --topic "company secrets" --technique 1 --output secret.txt

# Run end-to-end automated plant-and-query RAG poisoning lifecycle
crucible poison-test rag --ingest-url http://api/ingest --query-url http://api/query --topic "finances"

# List active poisoning evaluation sessions
crucible poison-test list

# Check the status of a specific poisoning session
crucible poison-test status <session-id>

# List all 12 reference targets (6 vulnerable, 6 hardened)
crucible target list

# Start a specific reference target (e.g. sql_vulnerable) on port 9000
crucible target start --name sql_vulnerable --port 9000

# Spin up all 12 targets, run health & ground-truth validation, write JSON report
crucible target validate --output ground_truth_report.json

# .github/workflows/security.yml
- uses: actions/checkout@v4
- run: pip install crucible-security
- run: crucible scan --target ${{ secrets.AGENT_URL }} --fail-on CRITICAL

- name: Crucible Security Scan
  uses: crucible-security/[email protected]
  with:
    target: ${{ secrets.AGENT_URL }}
    format_preset: openai
    model: gpt-4o
    headers: '{"Authorization": "Bearer ${{ secrets.OPENAI_API_KEY }}"}'
    fail_on_grade: C # Fails workflow if grade is C, D, or F

crucible/
  models.py                    # Pydantic data models
  cli.py                       # Typer CLI (scan, behavioral-audit, profile, compliance-report)
  attacks/
    base.py                    # BaseAttack ABC
    prompt_injection.py        # 50 attack vectors
    goal_hijacking.py          # 20 attack vectors
    jailbreaks.py              # 20 attack vectors
    enterprise_graph.py        # Cross-agent trust attacks
    memory_poisoning.py        # Persistent state attacks
    behavioral_escalation.py   # Multi-turn escalation sequences (v0.3)
    multi_turn_strategies.py   # Crescendo & Context Confusion (v0.3)
    profile_templates/         # Agent type detection templates (v0.3)
    multi_agent_contagion.py   # Cross-agent trust attacks (v0.4)
    dynamic_generator.py       # Research-driven attack gen (v0.4)
    hallucination.py           # 15 hallucination/overreliance attacks (v0.5)
    toxicity.py                # 20 toxicity/safety attacks (v0.5)
  modules/
    base.py                    # BaseModule ABC
    security.py                # Module registry
  core/
    runner.py                  # Async parallel scan engine (anyio)
    scorer.py                  # Deduction-based scoring + grading
    mutation_engine.py         # Payload obfuscation (6 strategies)
    behavioral_engine.py       # Multi-turn behavioral drift engine (v0.3)
    multi_turn_engine.py       # Multi-turn attack runner (v0.3)
    profiler.py                # Agent capability profiler (v0.3)
    compliance_engine.py       # EU AI Act mapping engine (v0.3)
    reporter.py                # Bug bounty report generator
    cache.py                   # TTL-based scan result cache
    research_engine.py         # Autonomous research orchestrator (v0.4)
    patcher.py                 # Auto-remediation engine (v0.4)
    canary.py                  # Active deception canaries (v0.4)
    statistics.py              # Zero-dependency bootstrap confidence engine (v0.6.1)
  reporters/
    base.py                    # BaseReporter ABC
    terminal.py                # Rich terminal renderer
    json_reporter.py           # JSON file exporter
    html_reporter.py           # Interactive HTML report
    slack.py                   # Slack webhook reporter
    compliance_reporter.py     # Compliance Markdown/JSON reporter (v0.3)
    huntr_reporter.py          # Bug bounty submission reporter (v0.4)
    sarif_reporter.py          # Export results to SARIF 2.1.0 (v0.5)
    atlas_reporter.py          # MITRE ATLAS compliance mapper (v0.6)
    nist_reporter.py           # NIST AI RMF compliance mapper (v0.6)
  poison/                      # Stateful memory & RAG poisoning package (v0.8.0)
    session_store.py           # Atomic JSON poisoning session store
    document_generator.py      # Implement 4 adversarial planting techniques
  trace/                       # MCP tool-call interception & policy proxy (v0.7.0)
    models.py                  # Pydantic trace models
    policy.py                  # YAML rule-based evaluation engine
    audit_log.py               # Append-only thread-safe JSONL logger
    proxy.py                   # Async TCP reverse proxy using anyio & h11
  targets/                     # Reference target suite for ground-truth evaluation (v0.18.0)
    base_target.py             # Abstract base HTTP target using Python standard library
    registry.py                # Central target registry mapping names to classes
    runner.py                  # Context-manager for starting and stopping targets cleanly

✗ Preflight failed: Target returned 405 Method Not Allowed.
  You specified --method GET but this endpoint requires POST.
  Re-run without --method GET or use --skip-preflight to bypass this check.