RNPLens reads structure files and JSON supplied by its users and writes JSON, text and HTML reports. Please report any way in which an input file, identifier or label can
- write or overwrite a file outside the chosen output directory,
- overwrite an existing result (RNPLens is designed never to do this),
- inject markup or script into a generated HTML report, or
- crash a parser in a way that could be exploited,
by private vulnerability report through this repository's GitHub Security tab ("Report a vulnerability"),
not through a public issue. Include the RNPLens version (rnplens --version), the command, and a minimal input
that reproduces the problem.
Security fixes are made on the latest release of the 0.4 series (the toolkit rnplens 0.4.x and the optional
rnplens-assistant 0.1.x). Earlier release candidates (0.2.0rcN, 0.4.0-rcN) are not supported.
The optional Assistant runs a language-model process inside an operating-system jail (bubblewrap on Linux, Seatbelt on macOS). Its confinement is not yet security-qualified (assistant/README.md). Please also report, privately, any way in which that process can read files, reach network destinations or receive environment variables beyond what docs/setup_and_privacy.md states.
The test suite includes hostile identifiers and file names (HTML escaping of Report and Diagnose pages), case identifiers that are paths (refused), and refusal to write into existing output directories.