接入你的智能体
發布者自己的指令。不是 PlugMCP 端點。
用我們的金鑰與日誌託管一份副本。
npx -y [email protected]Claude Code 沒有本機指令的安裝連結。複製下方指令並在終端機中執行。
claude mcp add shinthink-blitzstrike -- npx -y blitzstrike@2.4.198在終端機中執行。
{ "mcpServers": { "shinthink-blitzstrike": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}claude_desktop_config.json
遠端 URL 還是本機指令? 兩種安裝方式的差異
Reconnaissance at speed. Analysis in depth. Validation before report.
Blitz Strike is a structured penetration-testing methodology — reconnaissance,
source analysis, and validation — delivered as a universal MCP server. It
enumerates the attack surface (BLITZ), traces source-to-sink reachability
(EAGLE-EYE), and verifies each finding live before it is reported (STRIKE).
One server, every agent: scope enforcement to submission-ready findings in a
single run_engagement call, with the relevant exploit-tool manual attached to
every result.
A scan hit is a hypothesis. A live test is the verdict.
Blitz Strike exists to eliminate the two most common failure modes in automated security assessment: false positives from surface-level pattern matching, and unverified findings reported without live confirmation.
Blitz Strike is a Model Context Protocol (MCP) server (TypeScript / Bun) that
packages a 3-tier security-audit methodology as callable tools — and runs the
whole engagement server-side, so a single run_engagement call works from
Claude Code, Cursor, Hermes, OpenCode, Claude Desktop, Gemini, or any MCP client.
Blitz Strike maps a structured penetration-testing methodology — reconnaissance, source analysis, and validation — into three tool tiers executed server-side.
Reconnaissance → analysis → validation. Nothing is reported until STRIKE confirms it.
Beyond the three tiers, Blitz Strike also ships Web3 audit (deterministic Solidity + executable Foundry proof), a live logic-bug prober (sibling enumeration + differential IDOR), hunting intel (CWE/CVE watchlist + DNS sinkhole detection), and out-of-band proof — so IDOR/auth logic bugs, blind injection, and smart-contract findings are all caught and proven, not just pattern-matched.
Blitz Strike is driven by the LLM — Claude, Hermes, OpenCode, Codex, or any MCP client. The LLM is the brain (plans, routes, delegates, judges); Blitz Strike is the deterministic hands + knowledge + guardrails.
A full engagement is one call, or a granular agent-orchestrated cycle:
npx blitzstrike serve --mcp # connect your agent, then ask it to
# "audit ./src" (source) or "audit https://example.com" (live)
The LLM classifies the target automatically (URL → live pipeline, filesystem
path → source pipeline), then drives recon → analyze → verify → review →
report — guided by the bundled doctrine (instructions + skills + per-step
next_steps) and fanned out across the platform's native sub-agents.
→ Autonomy & doctrine — how the LLM is steered.
bun build --compile — ship one executable per platform.bunx blitzstrike / npx blitzstrike.@modelcontextprotocol/sdk).# Zero-install — works from any MCP client, no clone, no build
npx -y blitzstrike doctor # verify the environment
npx -y blitzstrike install # auto-register with every detected agent CLI
npx blitzstrike install detects every installed agent CLI (Claude Code,
Cursor, OpenCode, Codex, Hermes, Gemini, Windsurf, Copilot, Cline) and writes
the correct MCP config to each one in its native format. Restart your agent and
call run_engagement.
From source:
git clone https://github.com/shinthink/blitzstrike.git
cd blitzstrike
bun install
bun run src/index.ts serve --mcp
{
"mcpServers": {
"blitzstrike": {
"command": "blitzstrike",
"args": ["serve", "--mcp"]
}
}
}
claude_desktop_config.json or .mcp.json.cursor/mcp.json.mcp.jsonmcp_servers: in config.yamlRun blitzstrike install to print the exact snippet.
blitzstrike serve --mcp # start MCP server over stdio (default)
blitzstrike doctor # health check: runtime + 130-tool catalog + creds
blitzstrike install # write MCP config to detected clients (Claude/Cursor/OpenCode)
blitzstrike install --dry-run # preview the config without writing
blitzstrike sync-data # fetch heavy datasets (payloads + templates) on-demand
blitzstrike update # check for a newer version + refresh the data cache
blitzstrike version # print version
Each issue carries a fix: line — no guessing.
blitzstrike install detects which MCP client config files already exist
(Claude ~/.claude.json, Cursor ~/.cursor/mcp.json, project .mcp.json) and
merges the Blitz Strike server entry in — it never overwrites your existing
MCP servers. With no client detected, it prints the snippet for manual paste.
The intelligence layer (WAF signatures, tech/CVE/port correlations, fuzzer
data, vuln ontology, exploit payloads, nuclei detection templates) is sourced
from airecon (MIT), PayloadsAllTheThings (MIT), and nuclei-templates (MIT) —
loaded at runtime and wired into the tool surface above. The attack-vector
taxonomy (intelligence/attack_vectors.json) is first-party.
317 indexed tool manuals + 17 playbooks from kali-pentest (Apache-2.0) — 334 manual files, 100% of the 130-tool catalog documented. These are NOT decoration — they are wired into the flow:
tool_lookup(name) auto-attaches the tool's full manual.run_engagement() attaches the relevant manual per matched chain's tools field.Memory is append-only JSONL at ~/.blitzstrike/memory.jsonl (override with
BLITZSTRIKE_HOME). run_engagement auto-captures matched escalation chains
as pattern entries (deduped by chain id), so the knowledge base grows with
every engagement — no duplicate spam, and only verified=true entries are
authoritative.
130 self-written security tools (not copied from any project), each with command base, key flags, install command per platform, check_installed probe, phase, tags, alternatives, requires_root, pipes, and homepage.
32 skill playbooks (markdown) — universal, license-safe hidden gems from the internet (MIT). Our personal attack-tree playbooks were removed (they were private methodology, not universal exploit scripts).
ap-* prefix. Rare offensive playbooks:
cross-forest-trust-abuse, gitea-ci-injection, kerberos-trust-abuse,
multi-domain-ad-attacks, client-side-crypto-forgery.hp-* prefix. End-to-end audit playbooks:
full-security-audit, smart-contract-audit, api-security-test, sast-code-review,
container-image-scan, iac-cloud-posture.See ATTRIBUTION.md for full license/copyright notices.
58 data-driven escalation chains, each with ordered steps carrying:
tool_hint — which Blitz Strike tool to usesuccess_criteria — binary observable for the stepinvariant_check — the assumption that MUST hold for exploitationnegative_control — how to refute the findingExamples: ssrf_cloud_metadata, lfi_log_poison_rce, appkey_leak_deserialization_rce,
hmac_empty_key_forgery, intval_form_id_bypass, extract_variable_injection_lfi,
split_controller_upload_bypass, race_condition_double_spend, ssti_template_injection_rce,
jwt_alg_confusion_forgery, cache_poisoning_xss, subdomain_takeover, and more.
Edit chains.json to add knowledge — never hardcode in source.
All other tools need no credentials.
bun install # deps
bun run typecheck # tsc --noEmit
bun run build # ESM bundle → dist/index.js
bun run test # regression + benchmark suite (453 checks)
bun run compile # single static binary → dist/blitzstrike
Blitz Strike carries a benchmark suite and a coverage matrix so quality is a number you can re-run, not a claim.
run_enterprise_benchmark) — a labelled corpus of vulnerable +
safe fixtures across PHP, JavaScript, Python, and Java. Current result:
detection rate 1.0, false-positive rate 0, precision 1.0.coverage_matrix) — 5 languages × sink classes = 84%
coverage (php, javascript, python, java, rust). Uncovered pairs are
legitimately absent from the language (e.g. Java has no eval, Python/Java no
PHP-style file inclusion, JS no deserialization sink).list_attack_vectors / attack_vectors) — a
master reference of 34 attack-vector categories (588 vectors) so a driving
agent can map the entire attack surface, not just the obvious sinks.bun run test runs 453 checks covering finding
lifecycle, evidence integrity, CVSS math, taint tracing, dedup, report
integrity, and benchmark invariants. CI runs it on every push.eagle_eye to confirm scope.FilesMatch .+\.ph(ar|p|tml)$ has a $ anchor — .php.jpg does not execute.strike_verify reflects your marker AND the chain's negative_control stays inert.This tool is provided for educational and authorized security research only. Do not use against systems without explicit permission from the owner.
Blitz Strike — a universal MCP security-audit toolbelt. Reconnaissance at speed. Analysis in depth. Validation before report.
{ "mcpServers": { "shinthink-blitzstrike": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式,而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Blitzstrike 透過 stdio 在本機執行。
上次檢查時間:2026-10-26。 Blitz Strike — a universal MCP security-audit toolbelt. Reconnaissance at speed. Analysis in depth. Validation before report.