raptix
Drive Raptix — an autonomous multi-agent red-team framework — over MCP to run authorized penetration tests and bug-bounty engagements end to end, then watch and steer them live from chat. Launch an engagement against a target, poll its transcript to narrate progress, send messages to refocus it, and pull findings as SARIF. Use when the user asks to run a pentest/red-team engagement, hunt a bug bounty, do recon, exploit/scan a host, web app, API, network, cloud, Active Directory, mobile app, or smart contract WITH Raptix — or to check/resume a running engagement or report what Raptix found. Triggers: run a raptix engagement, pentest this with raptix, bug bounty, recon this target, red team this, scan this host, resume the engagement, what did raptix find, raptix status. Do NOT use for ad-hoc local tool runs (running nmap/sqlmap/ffuf directly) when no Raptix server is involved — this drives the Raptix orchestrator, not raw tools.
- License 许可
- Apache-2.0
- 所属仓库
- Akapi895/raptix
- 最近更新
- 2026-11-06
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- raptix
- license
- Apache-2.0
- version
- 2.0.0
- metadata
- { "hermes": { "tags": [ "raptix", "red-teaming", "penetration-testing", "bug-bounty", "mcp", "autonomous-agents", "recon", "exploitation", "sarif" ], "related_skills": [ "pentest-recon", "offensive-reporting", "reconnaissance" ] }, "homepage": "https://github.com/Akapi895/raptix" }
- description
- Drive Raptix — an autonomous multi-agent red-team framework — over MCP to run authorized penetration tests and bug-bounty engagements end to end, then watch and steer them live from chat. Launch an engagement against a target, poll its transcript to narrate progress, send messages to refocus it, and pull findings as SARIF. Use when the user asks to run a pentest/red-team engagement, hunt a bug bounty, do recon, exploit/scan a host, web app, API, network, cloud, Active Directory, mobile app, or smart contract WITH Raptix — or to check/resume a running engagement or report what Raptix found. Triggers: run a raptix engagement, pentest this with raptix, bug bounty, recon this target, red team this, scan this host, resume the engagement, what did raptix find, raptix status. Do NOT use for ad-hoc local tool runs (running nmap/sqlmap/ffuf directly) when no Raptix server is involved — this drives the Raptix orchestrator, not raw tools.
02 资源文件 2
- 参考文档examples.md4.6 KB
- 参考文档reference.md6.0 KB
03 SKILL.md 正文
Raptix engagements (over MCP)
Drive Raptix — an autonomous multi-agent red-team framework — as if its CLI were in this chat. Run an authorized engagement end to end (recon → exploitation → post-exploitation → reporting) across web, API, network, Active Directory, cloud, mobile, smart-contract, and binary targets, then watch it progress and steer it as it runs.
You interact through the raptix_* MCP tools (listed below). The heavy
work runs inside the Raptix server; you are the operator at the console.
Mental model — read this first
- An engagement is a thread.
raptix_start_engagementreturns athread_id. That is the handle for every other tool — there are no run ids to track. - The orchestrator (
raptixgraph) builds an OPPLAN and delegates to specialist sub-agents (recon, exploit, postexploit, analyst, reverser, cloud_hunter, ad_operator, mobile_operator, …) via atask()tool. You watch that narrative and nudge it. - Engagements are long and asynchronous (minutes to hours).
startreturns immediately; you poll and narrate. Never block waiting for completion.
Authorization — non-negotiable
- Only start engagements against assets the user has explicitly confirmed are in scope. If scope is unclear or missing, ask before starting — do not guess a target.
- ALWAYS pass scope + rules of engagement in
instruction: name the in-scope hosts/domains/paths and the explicit out-of-scope items. The orchestrator enforces RoE on every tool call, but you are responsible for giving it correct scope. - Decline targets that are plainly not the user's to test.
Prerequisites (verify on first failure)
- A Raptix LangGraph server must be running and reachable
(
RAPTIX_API_URL, defaulthttp://localhost:2024). If a tool errors with a connection failure, tell the user to start it (langgraph devor the Docker stack) — don't retry blindly. - The
raptixMCP server must be registered and launched withRAPTIX_SKIP_BOOT=1(fast start). See the integration docs.
Tools
| Tool | Use it to | Key args | Returns (key fields) |
|---|---|---|---|
raptix_list_graphs |
see available graphs | — | [{graph_id, name}] |
raptix_list_engagements |
browse / resume | limit |
[{thread_id, engagement_name, status}] |
raptix_start_engagement |
launch | targets[], instruction, scan_mode, engagement_name? |
{thread_id, engagement_name, run_id, status} |
raptix_transcript |
watch the narrative | thread_id, after_index, limit |
{messages[], next_index, total, run_status} |
raptix_watch |
live sub-agent burst | thread_id, max_seconds, max_events |
{events[], run_status} |
raptix_send_message |
steer / answer / /model |
thread_id, message |
{run_id, status} |
raptix_engagement_state |
OPPLAN / scope / phase | thread_id |
{engagement_name, message_count, values} |
raptix_engagement_status |
run status + findings ready | thread_id, engagement_name? |
{status, findings_available} |
raptix_engagement_findings |
pull results | engagement_name, include_sarif? |
{available, result_count, level_counts, sarif?} |
raptix_cancel_engagement |
stop the run | thread_id |
text |
Full parameters, defaults, clamps, and return schemas are in
reference.md. Worked end-to-end runs are in
examples.md.
The core loop
- Pick a graph. Usually
raptix(full kill chain). Usereconfor recon-only,soundwavefor planning.raptix_list_graphs()if unsure. - Start.
raptix_start_engagement(targets=[…], instruction="In scope: …; Out of scope: …", scan_mode="standard"). Savethread_idandengagement_name. - Watch + narrate. Loop
raptix_transcript(thread_id, after_index=<previous next_index>); summarise only the NEW messages for the user (coordinator decisions,task(<specialist>)delegations, results). For a live burst useraptix_watch(thread_id). Checkraptix_engagement_status(thread_id, engagement_name); stop polling whenstatusis terminal orfindings_availableis true. - Steer when useful:
raptix_send_message(thread_id, "skip the staging host, focus on the API"), answer the coordinator, or switch models with/model anthropic/claude-opus-4-8. - Report. When findings exist:
raptix_engagement_findings(engagement_name, include_sarif=true)→ present severity, counts, and reproduction.raptix_engagement_statefor the OPPLAN/phase. - Resume later.
raptix_list_engagements()→ reuse anythread_id.
Polling cadence (phone / chat friendly)
- Don't spam tools. While running, poll the transcript every ~15–30s and give the user a 1–2 line update per poll, not raw dumps.
- Use the returned
next_indexas your cursor so each update covers only new activity. raptix_watchblocks up tomax_seconds(≤45) — use it for a quick live glimpse, not as your main loop.
Interpreting results (fast guide)
- transcript.messages:
roleis user/assistant/tool.tool_callsliketask(recon)means a specialist was dispatched.toolmessages carry results. - status:
pending/running= working;success= finished;error/timeout/interrupted= stopped (say why; offer resume/restart);none= no run yet. - findings:
available=false→ not persisted yet, keep polling.level_countsmaps SARIF level → count (error= critical/high,warning= medium,note= low). Passinclude_sarif=trueto mine reproduction details. - engagement_state.values: OPPLAN, objectives, scope, phase, working files.
Errors & recovery
- Connection failure → the Raptix server isn't up at
RAPTIX_API_URL. Ask the user to start it; don't loop. findings_available=falsefor a while → normal early on; keep watching the transcript and report progress.status=error/timeout→ read the last transcript messages for the cause, summarise it, and offer tosend_messagea fix or start fresh.- No active run on
watch/cancel→ the engagement is idle/finished; usetranscript/findingsinstead.
See reference.md and examples.md for depth.
04 同仓库更多 Skills
wps-pixie-dust
WPS Pixie-Dust offline nonce attack (reaver -K / pixiewps) and fallback online PIN brute (bully) to recover the AP's WPA PSK without capturing a handshake.
wpa3-sae
WPA3-SAE transition-mode downgrade (DragonShift), SSID Confusion CVE-2023-52424, Dragonblood side-channels, and SAE captive-portal credential recovery against WPA3-Personal networks.
wpa2-psk
WPA/WPA2-PSK handshake capture via targeted deauth + PMKID (no deauth required) + offline hashcat cracking. The most common consumer/SMB encryption mode in 2026.
wpa-enterprise-eap
WPA/WPA2/WPA3-Enterprise (802.1X/EAP) rogue-RADIUS evil-twin for MSCHAPv2 capture, GTC downgrade, and PEAP relay. MSCHAPv2 capture equals a NetNTLM hash — the primary wireless on-ramp to Active Directory.
krack-fragattacks
KRACK key-reinstallation (CVE-2017-13077..13082) and FragAttacks fragmentation/aggregation flaws (CVE-2020-24586..24588, CVE-2020-26139..26147) against legacy or embedded 802.11 supplicants with poor patch cadence.
evil-twin-karma
Evil-twin rogue AP with KARMA/Mana PNL-probe response, captive-portal credential capture, and post-association MITM for PSK/open networks. Distinct from wpa-enterprise-eap which targets 802.1X.