Security and publication policy¶
Report a vulnerability privately¶
Email the project maintainer at luke.repko@gmail.com with “rfSentry security report” in the subject. Include the affected component, version or commit, expected behavior, observed impact, and a minimal reproduction using synthetic data. Keep vulnerability details out of public issues. Remove tokens, private keys, receiver locations and deployment logs from attachments.
Publication policy¶
Keep private operator/deployment information out of source, examples, build outputs and public diagnostics unless its publication has been explicitly approved.
Review personal domains, hostnames and network addresses; callsigns and linked example identities; unapproved personal email addresses; local account and home/workspace paths; actual device serials and installation mappings; and identifying telemetry, run identifiers or infrastructure fingerprints. Review associations across fields, not just individual strings. Passwords, tokens, private keys, Wi-Fi credentials, cookies and captured private user data must never enter publication inputs.
Preserve contributor attribution, copyright, upstream credits and the maintainer’s approved public commit identity. Publication approval for CI metadata covers the reviewed runner names and labels, host/workspace hashes, browser path, capacity and version fields. Additional deployment details require their own review.
Public geography, official datasets, public project/upstream URLs, supported hardware specifications, synthetic protocol values and integrity hashes remain useful public data. Public keys/certificates are not private keys, but their identity fields still need publication review. Avoid blanket bans on addresses, emails, coordinates or identifiers.
Configuration and examples¶
Public profiles and fixtures must be deliberately synthetic, internally consistent and clearly distinguished from genuine qualification evidence. Use the documented example domains/addresses and synthetic hardware namespace. Do not replace a deployment dependency with fictional values to pass a check. Inspect its configuration authority first; stop for operator instructions if a required private value is not supplied through supported configuration.
Actual appliance identity belongs in /etc/rfsentry/appliance.conf or an
explicit private --appliance-config input. Existing hardware settings remain
in /etc; upgrades preserve them. Fresh hardware/recovery inputs use a reviewed
private --profile-dir, outside Git or under ignored config/local/. Workstation
settings remain in their documented per-user configuration. See
hardware profiles and
appliance identity.
Ignored files remain private working inputs; ignore rules do not untrack files
or remove history. Never place private configuration/audits under build/,
dist/, public assets, evidence upload directories or package staging roots.
Exact sensitive-value manifests, originals and rewrite maps belong outside
the repository and all publication roots, in owner-only private storage.
Checks and limits¶
Before committing, review the index privately:
./scripts/check-secrets.sh
python3 scripts/check-publication.py
python3 scripts/check-publication.py --staged
git diff --cached --check
The pre-commit hook checks index blobs. Public checks need no private inputs; they report that private exact matching was NOT RUN. Do not paste unredacted diffs, scanner inputs or diagnostic content into public logs or issues.
Publication checks document structural rules, exact-match input protection, generated-output checks and coverage limits. Secret scanning does not establish privacy cleanup. Current files do not prove clean history; a rewritten branch does not prove retirement of old tags, releases, Actions records, caches, clones or other copies. Passing these checks is not release qualification. History rewriting requires separate, explicit authorization; this policy grants no blanket permission.
Actual credential exposure¶
If an active credential is found, stop publication and notify the operator privately without echoing it. Identify the affected service and required revocation/rotation. Follow an authorized incident-response procedure; otherwise request the required operator action. Do not automatically rotate credentials, change services or claim resolution without verifying that remediation. Deleting text or rewriting history is not credential revocation.