AGENT SKILL

mantis-configure

Configures and validates Mantis pipeline environments, sandbox mechanisms, and AI models. Use to set up workflow.json, auto-detect host capabilities, switch between sandboxes (static-only, gvisor, microsandbox, gce), select AI models, and run fast 1-2s preflight tests. Don't use for scanning source code or running attack campaigns.

适用职业开发工程师
License 许可
Apache-2.0
所属仓库
google/mantis
最近更新
2026-11-15
Stars
★ 2.8k
Forks
237

01 前言 Front Matter

name
mantis-configure
description
Configures and validates Mantis pipeline environments, sandbox mechanisms, and AI models. Use to set up workflow.json, auto-detect host capabilities, switch between sandboxes (static-only, gvisor, microsandbox, gce), select AI models, and run fast 1-2s preflight tests. Don't use for scanning source code or running attack campaigns.

03 SKILL.md 正文

Pipeline Configurator (/mantis-configure)

System Goal

Environment and Model Configurator. Configures workflow.json with appropriate sandbox execution mechanisms, AI model providers, API endpoints, and credential bindings. Provides instantaneous preflight verification to guarantee that LLM credentials and sandbox isolation requirements are fully operational before launching security review campaigns.

Command Definition

  • Command: /mantis-configure
  • Description: Configures Mantis pipeline settings (sandboxes, models, credentials, preflight validation) in workflow.json.
  • Execution Command:
    python3 "${MANTIS_HOME:-/path/to/mantis}/reference/scripts/configure.py" [flags...]
    

Path Anchoring Requirement (CRITICAL): The configuration script resides within the Mantis installation directory at reference/scripts/configure.py. You MUST invoke this script via an absolute path or via $MANTIS_HOME. NEVER execute python3 reference/scripts/configure.py using a relative path inside audited target repositories.

  • CLI Options:
    • --sandbox / -s: Sandbox mechanism (static-only, gvisor, microsandbox, gce).
    • --model / -m: Default LLM model (e.g. gemini-3.8-flash, vertex_ai/claude-opus-5, vertex_ai/zai_org/glm-5.2-maas, openai/{MODEL_ID}).
    • --api-base: Custom endpoint URL for OpenAI-compatible LLM servers (e.g. http://localhost:8000/v1).
    • --reasoning-effort: Reasoning effort level (low, medium, high).
    • --timeout: LLM request timeout in seconds.
    • --project / -p: GCP Project ID (for GCE sandbox or Vertex AI routing).
    • --zone / -z: GCP Zone (e.g. us-central1-b).
    • --image / -i: Sandbox image name (e.g. mantis-sandbox-image or mantis-sandbox:latest).
    • --subnet: GCE Subnet name (e.g. mantis-isolated-subnet).
    • --workdir: Sandbox guest workdir (default: /workspace).
    • --workflow / -w: Path to workflow.json (defaults to auto-discovery).
    • --db / -d: Path to SQLite knowledge database (default: knowledge.db).
    • --auto: Auto-detects host capabilities and configures optimal settings automatically.
    • --save: Explicitly saves configuration changes to workflow.local.json.
    • --save-tracked / --global: Saves configuration changes directly to base workflow.json.
    • --interactive: Interactive step-by-step terminal wizard.
    • --test / --preflight: Executes fast (1-2s) static validation tests verifying LLM configuration format and sandbox readiness.
    • --probe / --probe-llm: Actively probes LLM reachability, provider credentials, and client dependencies with a minimal test prompt (test, max 256 tokens).
    • --show: Displays current configuration and diagnostic status.
    • --dry-run: Simulates configuration changes without modifying files.
    • --update-nodes: Updates all agent nodes in workflow.json to use the specified default model.
    • --json: Outputs configuration status and preflight diagnostics in JSON.

Supported Sandbox Mechanisms

Sandbox Description Isolation Level Requirements
static-only Static analysis only; reproducer and dynamic patching disabled. Zero Host Risk None (Always available)
gvisor Networkless OCI container executed under Google gVisor (runsc). Process & Kernel sandbox docker or podman with runsc registered
microsandbox Ephemeral Linux microVM with hardware virtualization. Virtual Machine /dev/kvm read/write access
gce Hardened ephemeral Google Compute Engine VM via IAP SSH tunnel. Cloud Hypervisor gcloud CLI, active GCP auth & project

Supported Model Providers

  1. Gemini Models (Google / Vertex AI):
    • gemini-3.8-flash, gemini-3.5-flash-lite
    • vertex_ai/gemini-3.8-flash, vertex_ai/gemini-3.5-flash-lite
  2. Claude Models (Vertex AI Model Garden):
    • vertex_ai/claude-opus-5
  3. MaaS & Open Source Models (Vertex Model Garden):
    • vertex_ai/zai_org/glm-5.2-maas
  4. Custom OpenAI-Compatible Endpoints:
    • openai/{MODEL_ID} or vertex_ai/openai/{MODEL_ID}
    • Supports custom --api-base (e.g. vLLM, Ollama, LiteLLM proxy), --reasoning-effort, and --timeout.

Common CLI Workflows

1. Fast Preflight Verification (~1s) & Active Reachability Probe

Check if LLM configuration and sandbox requirements are operational:

# Fast static validation (~1s):
python3 "$MANTIS_HOME/reference/scripts/configure.py" --test

# Active live reachability probe against LLM provider:
python3 "$MANTIS_HOME/reference/scripts/configure.py" --test --probe

2. Auto-Detect and Configure

Automatically inspect host capabilities (/dev/kvm, docker/runsc, gcloud) and select the best available sandbox:

python3 "$MANTIS_HOME/reference/scripts/configure.py" --auto

3. Switch to Static Analysis (Zero Dependencies)

python3 "$MANTIS_HOME/reference/scripts/configure.py" --sandbox static-only

4. Configure gVisor Container Sandbox

python3 "$MANTIS_HOME/reference/scripts/configure.py" --sandbox gvisor --image mantis-sandbox:latest

5. Configure GCE Ephemeral Cloud Sandbox

python3 "$MANTIS_HOME/reference/scripts/configure.py" --sandbox gce --project my-gcp-project --zone us-central1-b

6. Switch AI Model to Claude or Custom Endpoint

# Vertex AI Claude
python3 "$MANTIS_HOME/reference/scripts/configure.py" --model vertex_ai/claude-opus-5

# Custom Local vLLM / OpenAI server
python3 "$MANTIS_HOME/reference/scripts/configure.py" --model openai/my-model --api-base http://localhost:8000/v1

7. Interactive Configuration Wizard

python3 "$MANTIS_HOME/reference/scripts/configure.py" --interactive

Python API Reference

When invoked programmatically from Python:

from scripts.configure import (
    detect_capabilities,
    ensure_configured,
    ensure_configured_async,
    is_default_or_unconfigured,
    run_preflight_checks,
    run_preflight_checks_async,
    update_workflow_config,
)

# 1. Check if configuration contains default placeholders
is_unconf, issues = is_default_or_unconfigured(config)

# 2. Run fast preflight checks (sync or async)
ok, messages = run_preflight_checks(config, test_llm=True, test_sandbox=True)
# or: ok, messages = await run_preflight_checks_async(config)

# 3. Ensure configured (auto-resolves defaults if unconfigured)
valid_config = ensure_configured(auto=True)
# or: valid_config = await ensure_configured_async(auto=True)

Input/Output Contract

  • Reads:
    • workflow.json and optional workflow.local.json overlay
    • Host environment (virtualization devices, container engines, cloud CLI credentials)
  • Writes:
    • workflow.local.json (or workflow.json when --save-tracked is set)

04 同仓库更多 Skills

google
google/mantis

mantis-threat-model

Synthesizes trust boundaries, attack surfaces, and attacker profiles into a living threat model. Use as Stage B of the Knowledge Base generation process, reading architecture and entity definitions from the KB. Don't use for analyzing source code or extracting raw learnings from JSONL files.

★ 2.8k
google
google/mantis

mantis-summarize

Pre-processes the repository by generating security-focused summaries (mantis-summary.md) for each directory to make planning and research more efficient. Use when starting a review campaign to map the codebase before threat modeling and planning. Don't use for executing code reviews, writing test scripts, or patching code.

资料检索与研究
★ 2.8k
google
google/mantis

mantis-structural-index

Builds a content-addressed semantic-unit index from source code for structural context. Use when a pinned or live codebase is available and structural cross-reference data would improve research quality. Don't use for findings analysis, patching, or reporting.

开发工程师
★ 2.8k
google
google/mantis

mantis-review

Independently reviews findings and filters out false positives. Use when consolidated findings need validation against the actual source code. Don't use for reproducing crashes or patching code.

★ 2.8k
google
google/mantis

mantis-researcher

Audits production source code files based on the strategy in workspace/plan.json. Use when a review plan exists and you need to perform static analysis and deep-dive reviews of targeted files. Don't use for planning, deduplicating, or writing patches.

★ 2.8k
google
google/mantis

mantis-reproduce

Generates and runs crash reproducers to verify security flaws. Use when viable findings exist and you need to write and execute a script or payload to verify the crash. Don't use for code auditing or patching.

★ 2.8k