Cybermes logs a finding only after the agent runs a working proof

Zyrexnn's Cybermes, released as v3.3.0 on GitHub and updated on 5 September 2026, is an offensive-security framework designed to be operated by a coding agent rather than by a human at a shell. It includes a native Go MCP server, cybermes-mcp, that exposes more than ten security tools and context providers over JSON-RPC 2.0 to any editor that supports the Model Context Protocol, including Claude Desktop, Cursor, Windsurf and Kilo Code. The repository is Apache 2.0, has received 762 stars in the twenty days since it was created, and has been forked 133 times.

The proof-of-concept check

The maintainer's own README argues that giving a coding agent a target and asking for vulnerabilities is easy, and that keeping the false-positive rate low is the whole engineering problem. Cybermes tries to solve it in the framework rather than in the prompt.

Every finding has to pass a check the project calls the Zero-False-Positive PoC Gate. Before anything reaches the findings/ directory the agent must produce a standalone, non-destructive Python script under pocs/poc_<name>.py, plus a raw HTTP trace, and both must reproduce the issue. A finding without a running PoC does not get logged, so the part that writes the report checks the part that reasons about bugs.

The Go binaries built around subfinder, nuclei and sqlmap

The Go binaries do the routine work. smart_pipe pulls the most useful endpoints, status codes, and leaked secrets out of large, mostly useless fuzz and crawl output, and archives the rest to disk so the model does not see it. secret_scan searches the raw reconnaissance directory for credentials that got captured. search_knowledge promises sub-50 millisecond offline lookup across 200 standard-operating-procedure guides, PayloadsAllTheThings, and HackTricks datasets, so a query for a technique like JWT algorithm confusion is answered without a web fetch. aggregate_reports compiles the confirmed findings into SUMMARY.md, a metadata.json metrics file, a standalone HTML dashboard, and a PDF named per target.

Below that layer are well-known tools. subfinder, httpx, katana, ffuf, nuclei and sqlmap are already connected, so the agent is asked what to test and the framework supplies the tools. Every target creates its own directory under reports/<TARGET_SLUG>/, and the workspace isolation stops context from one engagement passing into the next.

Two ideas worth copying, even for teams already using Semgrep or Burp

The two design choices worth looking at beyond the marketing copy are the PoC gate and the stream filter. The check means a finding gets a report row only after the agent proves it: the agent that hallucinated the SQL injection has to write the script that exploits it, and if the script does not work the finding is dropped. The stream filter deals with the other cost, which is that a raw nuclei run against a large target can use up tens of thousands of context tokens telling the model about 404s, and a filter that runs before the output reaches the model lowers the bill and protects the quality of the reasoning.

Neither of those ideas is unique to Cybermes: the pipeline pattern of "recon streams, tools verify, agent narrates" is appearing in several open-source frameworks. What is new here is packaging it as an MCP server that any editor can call, licensing it under Apache 2.0, and providing cross-platform installers for Windows, Linux, macOS and Docker with a doctor.py script that checks the environment before the first scan.

Two limits are worth stating plainly. This is offensive security tooling and the disclaimer at the bottom of the README says the framework is for authorised bug bounty, authorised penetration testing and internal red teaming. And Cybermes brings its own tools and playbooks, but a security team already committed to Semgrep, CodeQL, Burp Suite Pro or an internal SAST platform is not obviously replaced by it. The aim here is to give an agent the same tools a red teamer would run, in addition to whatever platform the team already uses.

Source

• Zyrexnn/Cybermes on GitHub