crucible

   ██████╗██████╗ ██╗   ██╗ ██████╗██╗██████╗ ██╗     ███████╗
  ██╔════╝██╔══██╗██║   ██║██╔════╝██║██╔══██╗██║     ██╔════╝
  ██║     ██████╔╝██║   ██║██║     ██║██████╔╝██║     █████╗
  ██║     ██╔══██╗██║   ██║██║     ██║██╔══██╗██║     ██╔══╝
  ╚██████╗██║  ██║╚██████╔╝╚██████╗██║██████╔╝███████╗███████╗
   ╚═════╝╚═╝  ╚═╝ ╚═════╝  ╚═════╝╚═╝╚═════╝ ╚══════╝╚══════╝
  

pip install crucible-security

crucible init --target https://meu-agente.com/api/chat
crucible scan --target https://meu-agente.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 # Falha no workflow se a nota for C, D ou F

crucible/
  models.py                    # Modelos de dados Pydantic
  cli.py                       # CLI Typer (scan, behavioral-audit, profile, compliance-report)
  attacks/
    base.py                    # ABC BaseAttack
    prompt_injection.py        # 50 vetores de ataque
    goal_hijacking.py          # 20 vetores de ataque
    jailbreaks.py              # 20 vetores de ataque
    enterprise_graph.py        # Ataques de confiança entre agentes
    memory_poisoning.py        # Ataques de estado persistente
    behavioral_escalation.py   # Sequências de escalada multi-turn (v0.3)
    multi_turn_strategies.py   # Crescendo e Confusão de Contexto (v0.3)
    profile_templates/         # Modelos de detecção de tipo de agente (v0.3)
    multi_agent_contagion.py   # Ataques de confiança entre agentes (v0.4)
    dynamic_generator.py       # Geração de ataques orientada por pesquisa (v0.4)
    hallucination.py           # 15 ataques de alucinação/confiança excessiva (v0.5)
    toxicity.py                # 20 ataques de toxicidade/segurança (v0.5)
  modules/
    base.py                    # ABC BaseModule
    security.py                # Registro de módulos
  core/
    runner.py                  # Mecanismo de varredura paralela assíncrona (anyio)
    scorer.py                  # Pontuação baseada em dedução + classificação
    mutation_engine.py         # Ofuscação de carga (6 estratégias)
    behavioral_engine.py       # Mecanismo de desvio comportamental multi-turn (v0.3)
    multi_turn_engine.py       # Executor de ataques multi-turn (v0.3)
    profiler.py                # Perfilador de capacidade do agente (v0.3)
    compliance_engine.py       # Mecanismo de mapeamento EU AI Act (v0.3)
    reporter.py                # Gerador de relatórios de bug bounty
    cache.py                   # Cache de resultados de varredura baseado em TTL
    research_engine.py         # Orquestrador de pesquisa autônoma (v0.4)
    patcher.py                 # Mecanismo de auto-remediação (v0.4)
    canary.py                  # Canários de decepção ativa (v0.4)
    statistics.py              # Mecanismo de confiança bootstrap sem dependências (v0.6.1)
  reporters/
    base.py                    # ABC BaseReporter
    terminal.py                # Renderizador de terminal Rich
    json_reporter.py           # Exportador de arquivo JSON
    html_reporter.py           # Relatório HTML interativo
    slack.py                   # Reporter de webhook Slack
    compliance_reporter.py     # Reporter de conformidade Markdown/JSON (v0.3)
    huntr_reporter.py          # Reporter de submissão de bug bounty (v0.4)
    sarif_reporter.py          # Exportação de resultados para SARIF 2.1.0 (v0.5)
    atlas_reporter.py          # Mapeador de conformidade MITRE ATLAS (v0.6)
    nist_reporter.py           # Mapeador de conformidade NIST AI RMF (v0.6)
  poison/                      # Pacote de envenenamento de memória e RAG com estado (v0.8.0)
    session_store.py           # Armazenamento atômico de sessão de envenenamento em JSON
    document_generator.py      # Implementa 4 técnicas de plantio adversarial
  trace/                       # Proxy de interceptação de chamadas de ferramentas MCP e política (v0.7.0)
    models.py                  # Modelos de rastreio Pydantic
    policy.py                  # Mecanismo de avaliação baseado em regras YAML
    audit_log.py               # Logger JSONL append-only thread-safe
    proxy.py                   # Proxy reverso TCP assíncrono usando anyio e h11
  targets/                     # Conjunto de alvos de referência para avaliação de referência (v0.18.0)
    base_target.py             # Alvo HTTP base abstrato usando biblioteca padrão Python
    registry.py                # Registro central de alvos mapeando nomes para classes
    runner.py                  # Gerenciador de contexto para iniciar e parar alvos de forma limpa

✗ 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.