Lint-Owl is a taint-tracking static analyzer written in Rust whose result is the data-flow path from an untrusted source to a dangerous sink, not a flat list of warnings. It proves whether tainted input actually reaches a command, SQL, SSRF, path traversal, or deserialization sink in Python, JavaScript, TypeScript, and PHP, and prints the exact source-to-sink chain. Built for real bug-bounty and code-review work.
Live demo · MIT licensed · written in Rust
Built from scratch by Pavan Nallamothu (LinkedIn, GitHub).
cargo run -- scan examples/vuln.py # one file
cargo run -- scan path/to/your/repo # a whole tree of .py files
Exits non-zero when it finds anything, so it drops into CI or a pre-commit hook
(--exit-zero to override). Known sanitizers (int, shlex.quote, ...) clear taint,
so quoted or cast input is not flagged.
== examples/vuln.py ==
command-injection [critical]: tainted data reaches a command-injection sink
[source] line 4: host = request.args.get("host")
[flows] line 5: cmd = "ping -c 1 " + host
[sink] line 6: os.system(cmd)
1 tainted path(s) across 1 file(s)
The source starts inside handle, a function the file never calls, so this also
shows the analyzer walking each function body as its own scope, not just top-level
code.
JSON output for tooling: cargo run -- scan examples/vuln.py --json.
Linters flag local smells. Semgrep matches patterns but does not prove the input reaches the sink. CodeQL proves it but is heavy and needs a database and a query language. Lint-Owl is one small binary whose output is the tainted path itself, and it is built to be agent-native (an MCP tool an AI asks "does user input reach this exec").
cargo run -- serve --port 8080
Open the URL, paste code, and see each finding as a source-to-sink chain with severity
badges. Or POST directly: curl -s localhost:8080/scan -H 'content-type: application/json' -d '{"code":"..."}'.
cargo run -- mcp
claude mcp add lint-owl -- /path/to/lint-owl mcp
Tool lint_owl_scan takes {code} and returns the tainted paths, so an AI reviewing a
diff can ask "does user input reach a sink here" and get the chain back.
Scans .py, .js/.ts, and .php. Emits readable paths, JSON (--json), or SARIF
(--sarif) for GitHub code scanning. Add your framework's sources/sinks with --config.
Rust. Hand-written lexers and a shared recursive-descent parser per language, and a forward taint engine that tracks provenance so every finding carries its source-to-sink line chain. See DESIGN.md.
v0.3: Python, JavaScript, and PHP. Source-to-sink paths for command injection, SQL
injection, SSRF, path traversal, insecure deserialization, and more. Control-flow (loop)
and inter-procedural (cross-function) taint, sanitizer awareness, configurable
sources/sinks (--config), SARIF output (--sarif), directory/repo scanning, CI exit
codes, severity, CLI + HTTP API + console + MCP. Two independent audits fixed. Results are
candidate paths a human confirms.