A long-running security WeChat account turned a penetration-testing agent into "one chat box to test three platforms": 21 stars in 677 days, almost all from its own readers.
What it is A Chinese web console for AI-assisted penetration testing, aimed at consultancy security engineers. In one chat box it tests websites, WeChat mini programs and Android/iOS apps together, with a small model vetting every command.
Why it took off The author is an old account in security circles with an existing readership. On launch day it hit 134 stars through their own account and syndication across platforms. AI pentest agents were also a hot topic in China at the time, so peer accounts and news sites picked it up.
CHow you could ride this Don't build another general pentest agent. Copy the formula instead: wrap an existing agent framework in an industry-specific shell, pick local needs that English projects ignore (MLPS compliance, incident response, mini-programs/apps only), launch through your own account, build a WeChat group, and pair every release with an article.
- Data (updated daily) · Repo created 2026-10-16 · 696 stars · 7 days +125
GitHub · stars Data 2026-10-17 → 2026-11-08 (23 days, 22 observations). Points before 10-28 come from deeply / archive backfill, not every day has one
Full-Chain-Sentinel-System is an open-source Chinese pentest agent that the long-running Chinese security account "Zhigong Shanfang Lab" released in Oct 16. It tests Web, WeChat mini programs and Android/iOS apps from one web chat box. By Nov 6 it had 677 stars and 209 forks (GitHub daily figures). In a niche where top projects routinely hit tens of thousands of stars, that number is modest, but how it got there and what its users are doing with it is useful for anyone building an open-source tool in a vertical.
Midday on Oct 16, a CSDN article with 397 reads
On 2026 Oct 16 12:59 (Beijing time), one hour after the database was built, 「知攻善防实验室」 posted 《[全链哨兵]全场景渗透测试 Agent 工具,正式开源!》 on CSDN, using CSDN's one-click distribution to sync it to WeChat Official Accounts, Weibo, Zhihu, Juejin, Baijiahao, cnblogs, Douyin and Xiaohongshu (RedNote). As of 11-06, the CSDN version had only 397 reads.
The article's opening is candid:
I've been building an engineering-grade product for a while... The previous engineering version could do a website for about 2 yuan. Now I've found that the consultancy crowd still prefers a chat-box interface, so I built that and am open-sourcing it.
A later line reads: "The underlying agent framework is built on Pi, so I won't reinvent the wheel." By the next day (10-17) the repo had 134 stars.
An article with 397 reads brought in 134 stars on day one, but it wasn't the article. It was the person behind the account. "Zhigong Shanfang Lab" is known for its free "emergency response target machine" series, and its walkthroughs are everywhere on Zhihu's range compilations, Alibaba Cloud's developer community and cnblogs. Its older GitHub projects, DeepSeekSelfTool (671 stars) and BlueTeamTools (564 stars), together have over sixteen hundred. Full-Chain Sentinel's first-day numbers were in the same range as those older projects.
What it is
Consultancy security engineers handle three kinds of targets day to day: websites, WeChat mini-programs and mobile apps. The author's launch post argues that the three front ends differ only in their carrier, and what's tested is the same backend server, so they can all go on one platform.
Full-Chain Sentinel is a web console plus an AI agent:
- Web: assign tasks directly in the chat box.
- WeChat mini-programs: decompile the mini-program yourself with tools and upload it as an attachment.
- Apps: prepare a rooted Android device (a VM works too), install the companion app, and reverse-connect it to the main program. Check "Android" in the chat box to load the Android-specific tools and prompts.
- Gatekeeper audit: a second small model (the author recommends Qwen's qwen-flash) watches every command the large model issues, "avoiding production impact as much as possible." Manual approval is on by default.
- Domestic-style companions: domestic asset-mapping sources like FOFA, Quake and Hunter; a Nuclei PoC subscription (the author specifically mentions frameworks like RuoYi and jeecg-boot); DingTalk notifications; CVSS vulnerability summaries; Markdown reports.
It ships as multi-platform binaries plus Docker, and the author encourages running it on a Linux server "running around the clock." MIT license. In 21 days there were 283 commits, up to v2.2.13.
How the stars came
| Date | What happened | stars (GitHub daily count) |
|---|---|---|
| 10-16 | Repo created; author's account launches it and syndicates across platforms | — |
| 10-17 | 134 | |
| 10-20 | Biggest single day in the observation window, +65; likely a repost by some WeChat account, though I couldn't identify which one | 259 |
| 10-22 to 10-28 | Picked up by peer WeChat accounts (tool recommendations), news sites like cn-sec, and Telegram security push channels; 10-26 also a CSDN "hands-on + 6 open-source tools compared" rewrite | 373 → 523 |
| 10-29 to 11-03 | No new moves, +4 to +23 per day | 528 → 580 |
| 11-03 | The author published another article, "In the age of automated pentesting, opening Burp again is just inelegant...", and shipped v2.2.12 and v2.2.13 at the same time | 580 |
| 11-04 to 11-06 | 637 → 670 → 677 |
A few details:
- All within the Chinese-language circle. Over the past 30 days, searches for "Full-Chain Sentinel", "Full-Chain-Sentinel" and "zgsfsec" on X each returned 0 results. Hacker News and Reddit have nothing, and there's no media coverage.
- The README includes two WeChat group QR codes, asking people to add a note saying "Full-Chain Sentinel". Once people join the group, there's a place to announce the next release or article.
- It's slowing down. In the last 7 days it gained +136, and in the first 7 days +168. Each new article or release brings a small bump again.
- Lots of forks, but not real forking-to-modify. The fork-to-star ratio is about 31%, versus roughly 11% typical for new projects of the same age. The top forks are all from repos with 0 to 1 stars and are unmodified, which looks more like the Chinese security community's habit of forking a copy to keep it from being deleted.
Is anyone actually using it?
Yes, but not many. The 11-04 v2.2.13 release was downloaded 688 times through 11-06 (Linux 305, Windows 197, Mac M-series 73, Android companion app 55), and the 11-03 v2.2.12 release was downloaded 487 times. Across both versions, nearly 1,200 downloads in three days is a lot for a repo with 677 stars, which suggests a meaningful share of the people who starred it actually installed it. There are only 4 issues: can asset mapping take a custom address, can other plugins be installed, one bug, and a failed call to the gatekeeper model.
Who it's competing with
| Competitors | Scale (GitHub stars) | Compared with Full-Chain Sentinel |
|---|---|---|
| Strix | 74k, +1,193 a week | English-language, aimed at R&D pipelines, no mini-program or app testing |
| Shannon | 51k | White-box: reads source code first, then attacks Web/API |
| PentAGI | 33k | English-language, multi-agent, fully autonomous |
| HexStrike AI | 13k | A bundle of MCP tools attached to existing AI clients, no console of its own |
| CyberStrikeAI | 8.4K | Chinese Go-based platform with an enterprise edition, mainly Web |
| ARTEX | Over 4.4 thousand stars in two weeks, plus several other reposted repos | A personal project from someone at ThreatBook, fully autonomous and targeting enterprise attack surfaces; got more attention after Microsoft named it in a report |
| VulnClaw | 3.9K | Chinese, MCP + Skill orchestration |
| BreachWeave | 727 | Also built on pi, multi-role, no mobile testing |
| wxmini-security-audit | 416 | A Claude Code skill that only audits WeChat mini-programs |
| Commercial products from Dbappsecurity, JD Cloud, ThreatBook and AWS | — | Paid, sold to the client side (Party A) |
This space is already crowded, and by star count Full-Chain Sentinel doesn't make the top ten. ARTEX, which came out the same month, has about 6 times its stars over two weeks. Full-Chain Sentinel occupies a narrow spot: security engineers who work in Chinese, deliver consultancy work, and need to test Web, mini-programs and apps. The top English projects and the Chinese ARTEX both skip mini-programs and real-device app reverse connection.
Why this one
The author has an existing readership. That's the main source of stars. Day one brought 134 stars and three weeks brought 677, the same range as the author's older projects. It never broke out of the niche.
Timing. In the second half of 2026, "AI pentest agents" were hot in China: Tencent Cloud ran a smart-pentest hackathon, ARTEX and CyberStrikeAI appeared, and "security agent comparison" articles came out in bulk. Readers were looking for a Chinese-language version they could install directly.
The angle targets consultancy day-to-day work. Testing all three platforms together, domestic asset-mapping sources, PoCs for domestic frameworks like RuoYi, DingTalk notifications, and reports you can hand over: English projects don't do any of this. That's what determines not how many people star it, but how many of the people who show up actually download it and join the group.
It addresses the fear of AI breaking a client's production environment. Consultancy engineers won't use a fully autonomous agent for fear of an incident they'd be blamed for. Default manual approval plus a small model watching every command speaks directly to that worry.
If it were you
C. The approach is worth copying, but this position isn't worth chasing. Full-Chain Sentinel's stars came mostly from readers the author built up over years, not from the product itself. The pentest-agent space has tens of thousands of stars at the top and several Chinese projects above a thousand, so a new "Chinese pentest agent" is hard to notice.
Who it's for: people who already have a readership on a security WeChat account, Bilibili or CSDN; or developers who have their own circle in a specific vertical (not just security) and know what that industry's people need to deliver.
Who it's not for: anyone with no distribution channel who expects a new general-purpose pentest agent to rack up thousands of GitHub stars. You'd be head-to-head with Strix and ARTEX.
How to do it: copy the playbook and swap the industry. The formula: take an existing agent framework (this one uses pi), wrap it in an industry shell (system prompts for that field, localized toolchains, a web console, human approval), pick a local need the English top projects skip, and you can ship something usable in a few weeks. Directions that fit: MLPS compliance testing and self-assessment, incident response and log forensics (the author's own background, and nobody is doing it specifically), deep versions for mini-programs only or apps only, alert triage for blue teams on the client side. Or fill a gap it leaves open: it needs a rooted Android device, and the iOS part is barely covered in the launch post, so a lower-barrier mobile testing agent has room.
How to get users: follow its sequence. On launch day, publish an article on your own account explaining "why I built it and how to use it", distribute it in one click to WeChat Official Accounts, Zhihu, Juejin and CSDN; put a WeChat group QR code in the README to pull people into a group; then pair every release with an article, each time pulling in another wave. Peer WeChat accounts' "tool recommendation" columns and Telegram security push channels will repost it for free. X and Hacker News barely help with this category. Without your own account, first spend months writing tutorials and building range labs to build readership. That is where Full-Chain Sentinel's real starting point was.
When to do it by: the domestic AI pentest agent wave is still hot and comparison articles are still coming out one after another. Once the leading Chinese projects fill in mini-programs, apps and compliance, the narrow niches shrink. The sooner the better.
Signs you've misread it: (1) if Full-Chain Sentinel passes several thousand stars in two or three months and spreads beyond the Chinese-speaking circle, that means the "full-chain across three platforms + gatekeeping" product point is itself attractive and worth an upgrade to B; following it still has a chance. (2) If the author launches a paid, engineered version and the open-source one becomes the funnel, the real value of this position lies in the commercial edition, not stars. (3) If your industry shell has been out for two weeks and few people download it or join the group, that field lacks enough people who need to ship something every day. Change direction.
---
Data and sources: GitHub repo, daily star/fork counts (10-17 to 11-06), release package downloads and Issues (11-06); the author's first CSDN article (10-16) and 11-03 newer articles; reposts by peer WeChat accounts and info sites, CSDN comparison articles (10-26), Telegram security push channel; X search over the past 30 days returned 0 (11-06); competitor star counts come from GitHub (around 11-06). Places marked "inference" are estimates.
Research notes: competitor table, acquisition evidence, full research report
- Competitors
Name Where Launched Scale How it differs from this one usestrix/strix Open source About 2025-08 73.9k stars English, aimed at R&D CI/CD find-and-fix; no mini-programs/apps KeygraphHQ/shannon Open source About 2025-09 51.3k stars white-box source reading, then Web/API attacks vxcontrol/pentagi Open source About 2024-12 33.4k stars English, fully autonomous multi-agent; no domestic scenarios 0x4m4/hexstrike-ai Open source About 2025-07 13.4k stars MCP toolkit, no own console or approval flow AIPentest/CyberStrikeAI Open source About 2025-11 8.4k stars Chinese Go platform + enterprise edition, mainly Web mhtsec/ARTEX (plus translation/repost repos) Open source 2026-10 first 4.4k stars in two weeks, several repost repos with thousands more a personal project from someone in the ThreatBook orbit, fully autonomous attack-surface testing for enterprises, with big-company backing and event-driven buzz Netw0rkNoob/VulnClaw Open source About 2026-04 3.9k stars OpenClaw + MCP/Skill orchestration, no mobile m-sec-org/BreachWeave Open source About 2026-04 727 stars also built on pi-mono, multi-role architecture, no mini-programs/apps Dbappsecurity / JD Cloud AIPTS / ThreatBook RedAgents / AWS Security Agent Big-company features 2025–2026 Commercial products paid SaaS / private deployment, aimed at enterprise security teams - Where users come from
- Verified author's own launch post (CSDN + Moyan one-click sync to WeChat Official Account/Zhihu/Weibo/Xiaohongshu, etc.): on the day the repo was created, 10-16 the CSDN launch article was published, and the page shows it was synced to multiple platforms; about 134 stars on day one
- Confirmed the author's existing brand (readers of the incident-response range series): Zhihu, Alibaba Cloud and cnblogs cite its ranges extensively; older repo DeepSeekSelfTool has 671 stars, BlueTeamTools 564 stars
- Confirmed reposts by peer WeChat accounts / info sites / marketing accounts: xtcaq.com tool recommendation, cn-sec.com, a CSDN marketing-account comparison (10-26), Telegram cvebird push, agenttape listing
- Confirmed README WeChat group funnel: two WeChat group QR codes on the homepage, with a request to note "Full-Chain Sentinel" when joining
- Confirmed sustained posting + frequent releases: 11-03 new article, v2.2.12/2.2.13,11-04 released the same day, +57 stars
- Inference X/HN/Reddit: x-radar returned 0 results over the past 30 days
Why Full-Chain Sentinel (Full-Chain-Sentinel-System, by zgsfsec) got traction
Source: GitHub · repo created 2026-10-16 · 2026-11-06:677 stars as of the database snapshot / 691 stars live via GraphQL, 219 forks · research date 2026-11-06
Bottom line
This repo didn't earn its stars through novel tech. It was launched by a veteran Chinese security content creator, "Zhigong Shanfang Lab" (WeChat Official Account + CSDN, known for its free incident-response range series), on the creator's own accounts: a Chinese web console for a penetration-testing agent aimed at commercial security service delivery, built on the open-source agent framework pi, with the pitch that "Web + WeChat mini-programs + Android/iOS apps all run from one chat box." The stars come from the author's existing followers, plus reposts from Chinese security info sites and marketing accounts. The circle is small and the ceiling is low. In the same period, ARTEX (ThreatBook-affiliated) passed 4k+ stars in two weeks, so by comparison this is a mid-sized "in-circle" project.
1. Data
- Star curve (GitHub repo_day / data card): 10-17 134 → 10-20 259 (+65 that day, the largest jump in the observation window) → 10-22 373 → 10-28 523 → 11-03 580 → 11-04 637 (+57, in sync with the v2.2.12/2.2.13 release and the author's 11-03 new article) → 11-06 677. gain_7d 136, first 7 days 168; growth is slowing, and each release or article gives it another bump.
- Fork-to-star ratio ≈ 31% (209/677), well above the ~11% typical for new repos (same-age median from the database schema). Chinese security tools often see people fork to save a copy in case the original is deleted; the top forks all have 0–1 stars and no changes, so these are not derivative works.
- Evidence of real use: download counts for the release packages in the last two versions (GitHub GraphQL, 2026-11-06): v2.2.13 (released 11-04) 688 downloads total (linux-x64 305, windows 197, mac arm 73, Android DroidBridge APK 55…); v2.2.12 (released 11-03) 487 downloads total. Older release assets have been deleted, so cumulative totals aren't visible. There are only 4 Issues (custom URLs for asset mapping, whether it can install "break-armor" tooling, bugs, gatekeeper model call failures), so people are installing and using it, but not at high volume.
- 283 commits, reaching v2.2.13 in 21 days, iterating very fast.
- Author: GitHub account zgsfsec = "Zhigong Shanfang Lab", 226 followers, account registered 2021 years ago. Older repos include DeepSeekSelfTool (671 stars), BlueTeamTools (564 stars), HuntBack (206 stars), and more, all built up by the same readership (GitHub GraphQL).
II. Competitors
| Product | Form | Launched | Scale | How it differs from Full-Chain Sentinel |
|---|---|---|---|---|
| usestrix/strix | Open source | About 2025-08 | 73.9k stars, +1193 this week | English, aimed at R&D CI/CD "find and fix"; no mini-programs/apps |
| KeygraphHQ/shannon | Open source | About 2025-09 | 51.3k stars | mainly white-box, reading source first, then Web/API |
| vxcontrol/pentagi | Open source | About 2024-12 | 33.4k stars | multi-agent, fully autonomous; English |
| PentestGPT | Open source | 2023 | about 14.9k stars (Zhihu comparison) | academic roots, leans toward CTF/ranges |
| 0x4m4/hexstrike-ai | Open source | About 2025-07 | 13.4k stars | MCP toolkit, plugs into existing AI clients, no console of its own |
| AIPentest/CyberStrikeAI | open source (domestic) | About 2025-11 | 8.4k stars | Chinese Go platform with an enterprise edition; mainly Web |
| mhtsec/ARTEX (plus several repost/translation repos) | open source (domestic) | Around 2026-10-08 | 4.4k stars (10-17, within two weeks), with BaaRai/ARTEX, jiwoochris/artex-global and others pulling thousands more | a personal project from the lead of ThreatBook's RedAgents, fully autonomous multi-agent attack-surface testing for enterprises; later named in a Microsoft report as used by state-level hackers, which added more buzz |
| Netw0rkNoob/VulnClaw | open source (domestic) | About 2026-04 | 3.9k stars | MCP + Skill orchestration built on OpenClaw |
| m-sec-org/BreachWeave | open source (domestic) | About 2026-04 | 727 stars | also built on pi-mono, multi-role pentest agent; no mobile |
| sssmmmwww/wxmini-security-audit | open source (domestic) | About 2026-04 | 416 stars | a Claude Code Skill for WeChat mini-program auditing only |
| Claude-Red / Anthropic-Cybersecurity-Skills / reverse-skill and others | Open-source skill pack | 2026 | 10k–44k stars | skill packs installed into Claude Code/Codex, not standalone platforms |
| Dbappsecurity, JD Cloud AIPTS, ThreatBook RedAgents, AWS Security Agent | big-vendor commercial products | 2025–2026 | Commercial | paid SaaS / private deployment, aimed at enterprise security teams |
Bottom line: the category is already crowded, and Full-Chain Sentinel doesn't crack the top ten by stars. It targets a narrow niche: Chinese commercial security engineers who need to test Web, mini-programs and apps every day.
3. Growth tactics (where users come from)
- Launched on the author's own account (confirmed): on the day the repo was created (10-16 12:59 Beijing time, about 1 hours after creation), CSDN published "[Full-Chain Sentinel] Full-scenario penetration-testing agent tool, now open source!" The page notes it was synced in one click via "Moyan" to WeChat Official Accounts, Weibo, Zhihu, Juejin, Baijiahao, cnblogs, Douyin and Xiaohongshu. It got about 134 stars on day one. Source: blog.csdn.net/Liyu_6618/article/details/168130447
- The author's existing brand and readers (confirmed): the incident-response range series from "Zhigong Shanfang Lab" is widely cited in Chinese security teaching circles (Zhihu "20+ range collection", Alibaba Cloud developer community, cnblogs, many personal blogs with writeups). The older repos total about 1,600+ stars. Most of the new repo's stars come from this readership.
- README WeChat group funnel (confirmed): two WeChat group QR codes on the homepage, with a request to note "Full-Chain Sentinel", to move users into a private community.
- Automatic reposts by security info aggregators / push channels (confirmed): the Telegram "Pentest/Security Push Center" (t.me/s/cvebird) pushed this repo; agenttape.com lists it; cn-sec.com reposted the comparison.
- Reposts by peer accounts and marketing accounts (confirmed): Shandong Xintianchen (xtcaq.com) "Tool recommendation | Full-Chain Sentinel: Web, mini-program and app penetration testing in one agent with gatekeeping audit"; a CSDN post by the account "小司机的奥拓" 10-26, "AI Pentesting from Beginner to Expert: Full-Chain Sentinel in Practice + 6 open-source solution comparison" (191 reads, resource pack link at the end, a typical marketing-account rewrite); 10-20 that day's +65 stars probably came from one of the WeChat reposts, but which one has not been verified.
- Sustained posting + releases (confirmed): 11-03 the author published "In the age of automated pentesting, opening Burp by hand just isn't elegant anymore…", and at the same time released v2.2.12/2.2.13,11-04, +57 stars that day.
- X / HN / Reddit (none): over the past 30 days, x-radar searches for "Full-Chain Sentinel", "Full-Chain-Sentinel" and "zgsfsec" each returned 0 results; almost nobody in the English-speaking community mentions it.
- On Bilibili, searching "全链哨兵" (Full-Chain Sentinel) turned up no relevant videos (only a game with the same name); one video, "ChinaRan404: 【全链哨兵】全链路渗透测试Agent 系统" 777 views, appears in the search snippets, unverified.
IV. Why it ran (by weight)
- Resources: the author is an old account with an existing readership. The day-one cold start was about 134 stars, about the scale of the author's older repos (DeepSeekSelfTool 671, BlueTeamTools 564); stars mostly stayed within the author's own circle.
- Timing: in the second half of 2026, "AI pentest agents" were hot in China: the Tencent Cloud intelligent pentest hackathon, ARTEX, CyberStrikeAI, and "12 security agent comparisons" all showed up in quick succession, and readers were looking for a "domestic, Chinese-language, ready-to-use" version.
- The product targets the everyday work of service providers: Web + WeChat Mini Program (upload decompiled packages) + reverse connection from rooted Android devices + iOS, all from one chat box; domestic intel sources (FOFA / Quake / Hunter), Nuclei POC subscriptions, DingTalk notifications, a CVSS vulnerability hub, Markdown reports. These are core needs for security delivery work, and English projects mostly skip them.
- Out-of-the-box distribution: multi-platform binaries + Docker, a Kali work container, hot-switching of proxy/WireGuard. Nearly 1,200 downloads in 3 days across the last two releases show people are actually downloading and running it, not just starring it.
- Controllability addresses service providers' concerns: human approval by default, plus a small model (recommended qwen-flash) that "guards the gate" by reviewing every Bash call, guarding against AI damaging a client's production environment. This is exactly why service engineers won't use fully autonomous agents.
5. Differentiation, and whether it's why it took off
The differentiation is "Chinese language + service-provider security scenarios + full chain across three ends (Web / Mini Program / APP) + approval gating", with the underlying layer not reinvented: it forks pi directly. Functionally, the leading English projects (Strix, PentAGI, Shannon) don't do Mini Programs or real-device APP reverse connections, and the domestic ARTEX and CyberStrikeAI focus on Web / enterprise attack surfaces. But this isn't the main source of stars: stars mainly come from the author's account and reposts. Differentiation determines how many of the people who arrive actually download and join groups (download counts and issue content show this), not how much exposure there is. ARTEX in the same niche has big-company backing and event heat, and over two weeks it got about 6 times as many stars.
6. Visibility and copycats
- Visibility: it's circulating in small circles. Domestic security WeChat Official Accounts / CSDN / news sites reposted it, and it was at 0 on X and HN; no media coverage. It's an open-source project, with no revenue involved.
- Copycats: so far there are no libraries with the same name or copied versions. The forks are all 0–1-star verbatim bookmarks; rewritten clickbait articles have already appeared (CSDN "小司机的奥拓"), but that counts as reposting, not cloning the product.
VII. What to take away
- Build a "vertical scenario shell" on an agent framework (pi): system prompts + localized toolchain + web console + approval. A usable version can ship in a few weeks.
- First pick local must-haves the leading English projects skip (WeChat Mini Programs, domestic mapping sources, DingTalk), and make them the selling points.
- Use "a small model auditing a large model" to reduce professional users' distrust of autonomous agents.
- Own accounts + one-click multi-platform distribution + README directing people into WeChat groups. Cold start and retention both happen in private channels.
VIII. Unknowns
- Which repost drove the increase of 10-20 (+65) and the growth around 10-24 is unverified; the read counts for the original WeChat Official Account articles weren't obtained (WeChat doesn't expose them).
- How much of the 31% high fork rate is "bookmark-to-keep-from-deletion" versus real modification was not checked individually (only the first 5 forks were reviewed).
- The author said on CSDN they previously made an "engineering version" (about ¥2 to test one website). Whether there's a commercial version, and whether the open-source version is funneling people to it, is unverified.
- Whether the stars were inflated: stargazer details weren't obtained (token-limited). Judging from the smooth curve and matching download volume, it doesn't look inflated.
Sources
- GitHub repo and README: github.com/zgsfsec/Full-Chain-Sentinel-System; GitHub GraphQL (stars/forks/releases downloads/issues/author repos, 2026-11-06)
- Our GitHub repo: repo_profile / repo_day (2026-11-06); competitor star counts also come from here
- Author's original post: blog.csdn.net/Liyu_6618/article/details/168130447; author's CSDN homepage blog.csdn.net/Liyu_6618
- Reposts: blog.csdn.net/web22050702/article/details/169147679; cn-sec.com/archives/5527032.html; www.xtcaq.com/h-col-104.html; t.me/s/cvebird; agenttape.com/agents/full-chain-sentinel-system
- Niche background: zhuanlan.zhihu.com/p/2073787149166628994; m.sohu.com/a/1090090800_104421; tsecbench.zc.tencent.com/resources
- X: x-radar search over the last 30 days gives 0 (2026-11-06)