Driga reads agent histories, process metadata, and local Git state. It can share collected data with paired machines and perform explicitly requested cleanup. Security reports are taken seriously and are welcome.

Do not open a public issue.

Report privately through either channel:

• GitHub private vulnerability reporting — preferred, keeps the report and the fix in one place.
• Email security@labstack.com. General open-source questions can go to opensource@labstack.com.

Please include what you can: affected version (driga version), a description of the issue, the impact you believe it has, and steps or scrubbed fixtures that reproduce it. A partial report is better than no report; do not wait until you have a full exploit. Do not include private transcripts or credentials.

We will acknowledge your report within 3 business days and give you an assessment and expected timeline within 10 business days. You will be kept informed as the fix progresses, and credited in the advisory unless you prefer otherwise.

Please give us a reasonable opportunity to release a fix before disclosing publicly. We will not pursue legal action against anyone acting in good faith under this policy.

There is no bug bounty program.

Driga assumes the operator's workstation and deliberately paired machines are trusted. In scope:

• Unintended disclosure of agent data or credentials through storage, logs, exports, the desktop app, MCP tools, or fleet sharing.
• Bypasses of fleet identity pinning, pairing, or per-viewer grants.
• Bypasses of the desktop app's content security policy, or of the check that its actions apply only to local sessions.
• Command injection or cleanup that acts outside the confirmed target.
• Process identity checks that allow signaling an unrelated process.
• Hook installation that overwrites unrelated agent configuration.

Out of scope:

• Vulnerabilities in the operating system, Git, or agent CLIs themselves.
• An operator deliberately performing an action they are authorized to perform.
• Anything requiring an already compromised workstation or malicious trusted peer.
• Secret scrubbing as a guarantee that arbitrary text contains no private data; reports of specific unintended captures or policy bypasses are welcome.

Driga is in alpha. Fixes land on the latest release only. There is no backporting to earlier versions and no long-term support branch.