Please report problems privately. Do not open a public issue.
- Use GitHub private vulnerability reporting: the "Report a vulnerability" button on the Security tab of
https://github.com/Prath2000/drashta. - Include the version (
drashta -version), steps to reproduce, and what you expect an attacker could do. - You can expect a first reply within about a week. This is a hobby-scale project, so there is no fixed fix time.
Only the latest commit on main is supported. There are no maintained release branches.
In scope:
- A web page (or other browser content) reaching the local server despite the Host, Origin, Sec-Fetch-Site and token checks.
- Running a command or reading or writing a file outside what the API should allow (see
internal/runxandinternal/pathguard). - Secrets from Claude Code transcripts reaching the index, search or API without redaction.
- The embedded-page and frame checks (
/api/servers/embed).
Out of scope:
- Anything that needs another process already running under your user account. That is trusted by design.
- Exposing the port to a network on purpose.
- Secrets that your own tools write to files drashta is asked to show.
- drashta binds to
127.0.0.1only. - Local processes under your user are trusted.
- Browsers are blocked by Host, Origin and
Sec-Fetch-Sitechecks. Mutating calls also need theX-Drashta-Tokenheader; the token is embedded inindex.html. - Commands run without a shell, through
internal/runx, with paths checked byinternal/pathguard. The editor command is set only inconfig.toml. - Transcript text is redacted before indexing. Redaction is pattern based and may miss a new secret format; reports of missed patterns are welcome.