Skip to content

Security: siddath/padipps

SECURITY.md

Security policy

Padipps is a local study tool with an optional loopback Codex adapter, not a public AI gateway. Run the backend only on your own machine and keep its loopback binding. The static Pages deployment contains no backend or credential store.

Report a vulnerability privately

Use GitHub private vulnerability reporting. Include the affected revision, reproducible steps, expected/actual behavior and a synthetic example. Do not put credentials, personal notebooks or exploitable details in a public issue. If private reporting is unavailable, ask for a private reporting channel in an issue without including the sensitive details.

Supported boundary

The current release uses a fixed Codex CLI version and model, explicit local opt-in, isolated startup context, disabled executable tools, strict event parsing and same-origin request validation. Unsupported CLI versions fail closed. A provider or platform update needs new isolation, cancellation and failure-path verification. Standard study features remain usable without an AI connection.

Enable GitHub secret scanning and push protection on forks. Use a GitHub noreply commit email unless you explicitly choose a public contact address. publication-policy.json permits only exact, owner-approved public contact addresses in commit metadata; adding someone else's address requires their consent. All credential, private-file and learner-data checks still apply. Before pushing, run the publication audit against files and reachable history, inspect the diff, and check that dist/ contains only allowlisted browser assets. A key added and later deleted can remain in history. Rotate a leaked credential through its issuing provider and coordinate history remediation; deleting its latest copy is not sufficient.

Verification limits

The audit covers relevant configuration, access control, injection, data-egress and exceptional-condition boundaries, informed by OWASP Top 10:2025, ASVS 5.0.0 and API Security Top 10:2023. Automated checks and sampled live calls are not a penetration test, a guarantee of secret-free future contributions, or compliance certification.

There aren't any published security advisories