Report privately. Do not open a public issue, pull request or discussion about it.

1. GitHub: open this repository's Security tab and choose "Report a vulnerability". This creates a private advisory that only the maintainers can see. If that option is not shown, private vulnerability reporting has not been enabled for this repository yet; use email.
2. Email: drake@aienos.com with the subject [pi-register] security report.

Include the commit or release you tested, the component (path), steps to reproduce, and what an attacker gains. If the report involves a secret, say where it was exposed and do not send the secret itself. Encrypted mail is not required; do not post proof-of-concept payloads publicly.

Reports receive best-effort review. No response deadline, fix deadline or embargo duration is promised here. Agree on disclosure and attribution with the maintainer through the private reporting channel before publishing sensitive details.

pi-register is a Rust CLI (src/main.rs) that reads a TOML service register and compares it with what systemd reports.

In scope:

• Register parsing (toml + serde) that accepts a malformed file silently, or a duplicate-name check that can be bypassed.
• A way for start, stop, enable or disable to act on a unit that is not in the register, or for a register value to be interpreted by systemctl as an option rather than a unit name.
• Register lookup that picks up a file the operator did not intend (the fallback order is --register <path>, then /etc/pi-register/register.toml, then ./register.toml). status and show run every health value through sh -c with the caller's privileges, so a register is code.
• A network call made by pi-register itself (AGENTS.md says it makes none).
• Secret values, private addresses, internal hosts or real inventories in tracked files.

Out of scope:

• The behaviour of a health command written by the operator, systemd, and the host operating system.

Only the default branch main; there is no release (Cargo version 0.1.0). Reports against older commits are welcome but fixes land on the default branch only.

An operations tool that supports Sovereign Intelligence (SI); it is not part of Sovereign Intelligence itself (AGENTS.md). There is no repository-defined build/test workflow; the build and test results documented here are local. GitHub-managed CodeQL analysis can appear on pull requests but does not run these tests. No certification, no bug bounty. See AUDIT_SCOPE.md for what to review and how to build and test it.