Introducing Vulnascan
Vulnascan is a deterministic, non-LLM static security scanner for your application’s local project directory. Provide it the directory containing your application code and get a detailed vulnerability report with remediation advice. No model, no randomness — identical report for identical input in every run.
- Finds vulnerable code (SAST). A pattern engine — regex + AST rules — searches your source for injection, XSS, path traversal, hardcoded secrets, weak crypto, insecure deserialization, SSRF, and around two dozen other vulnerability classes (28 rule types in all). Every finding carries a CWE id and a vulnerable → secure remediation pair.
- Finds vulnerable dependencies (SCA). It parses your lockfiles and checks every pinned package against the OSV vulnerability database (CVE / GHSA / OSV advisories, with the fixed version and severity). Reported in a separate section from the code findings.
- Produces a report you can act on.
text,markdown,json,sarif, or a single self-containedhtmlfile — each finding with its location, a masked code snippet, a plain-language explanation, and the fix. Secrets are masked before they ever reach a report.
The pipeline runs in fixed order — discover files → SAST rules
→ confidence gate → dependency scan (OSV) → CVE enrichment (NVD).
A network or parse failure never aborts the scan; it marks the report
PARTIAL and lists what was skipped, and the code findings are still
complete. --offline skips all network and still runs the full SAST pass.
OSV/NVD responses are cached on disk for 24 h.
Vulnascan is free and open source, and ships in three shapes over one shared engine.
All three call the same scan(path, ScanConfig) -> ScanResult —
there are no divergent code paths.
Desktop app
Double-click, choose a folder, read a rendered report. No Python, no terminal.
Read the desktop guide →CLI
security-preview scan <path> — for terminals and CI, with
--format, --offline, --min-confidence, and
exit-code gating.
Coding-agent skill
Claude Code or Cursor run the scan on request and summarise the findings from the JSON report.
Read the skill guide →Pattern rules match 14 languages and config formats (17 file extensions). Coverage is deep for the first six and targeted for the rest; dependency (CVE) scanning is separate and language-agnostic.
| Language | Extensions | Rule checks |
|---|---|---|
| Python | .py | 24 |
| JavaScript | .js .jsx | 24 |
| TypeScript | .ts .tsx | 24 |
| Java | .java | 24 |
| PHP | .php | 23 |
| Ruby | .rb | 17 |
| Go | .go | 14 |
| C# | .cs | 7 |
| YAML | .yml .yaml | 3 |
| JSON | .json | 3 |
| C | .c | 1 |
| C++ | .cpp | 1 |
| HTML | .html | 1 |
| Shell | .sh | 1 |
Dependency scanning reads lockfiles for Python, npm, Go, Ruby and Maven
(requirements.txt, poetry.lock, Pipfile.lock,
package-lock.json, yarn.lock, go.mod,
Gemfile.lock, pom.xml) and matches every pinned package
against the OSV database.