diff --git a/docs/operating-guidelines/operating-principles.md b/docs/operating-guidelines/operating-principles.md
new file mode 100644
index 0000000..cd3f873
--- /dev/null
+++ b/docs/operating-guidelines/operating-principles.md
@@ -0,0 +1,70 @@
+# Open Source Committee Operating Principles
+
+> These are the committee's day-to-day working rules of engagement. Per issue
+> #25, they live in the repo as version-controlled markdown at
+> `docs/operating-guidelines/operating-principles.md`, amended via pull request.
+>
+> Authority chain: these principles sit below the Committee Charter
+> (`committee-charter.md`, issue #9) and the Agentics Foundation mission and
+> vision above them, and above the mechanical OSS Intake Process below them. They
+> are meant to mature into testable definitions over time through tracked issues,
+> not to swell into a fixed document. Where these principles and the canonical
+> governance document differ, the governance document and the Board govern.
+
+## Purpose
+
+These principles guide how the Open Source Committee operates day-to-day. They
+sit below the Foundation mission and the committee charter, and above the
+mechanical intake process. They are meant to mature into testable definitions
+over time rather than to be a fixed charter.
+
+## Principles
+
+1. **Humans judge, agents execute.** (AF-GOV-002) Agents provide analysis,
+   logistics, case briefs, and infrastructure. Agents never cast votes, decide
+   escalation, or propose retraction. The committee's human judgment is the
+   point; automation makes that judgment efficient and well-informed.
+
+2. **Issues over documents.** Decisions, action items, and working artifacts live
+   as tracked GitHub issues with named owners, not as prose in meeting notes or
+   static docs. If it matters, it has an issue number.
+
+3. **The governance doc is the source of truth.** Where implementation diverges
+   from the canonical intake rules, the doc wins. Code changes follow doc
+   changes, not the reverse.
+
+4. **Repos are learning resources, not just libraries.** Given how fast code is
+   now generated, the committee's value is guidance and balances for the
+   ecosystem, including security and provenance, not raw volume.
+
+5. **Member-only, mission-aligned intake.** Submissions come from members in good
+   standing, are open source with a clear license, and advance the agentic AI
+   ecosystem. Incomplete or non-compliant submissions may be declined without
+   review.
+
+6. **Two-step, simple-majority voting.** Escalation vote, then validation vote;
+   a simple majority of members voting. Submitters do not vote on their own
+   submission or its retraction.
+
+7. **Tiered stewardship.** T1 Foundation-Owned, T2 Foundation-Endorsed (approved
+   through intake, listed on the website), T3 Community-Associated. Tier names
+   remain open for discussion (#10).
+
+8. **Auditability by default.** Every status change, score, and vote leaves a
+   durable trail in the issue and the attestation log. No decisions in DMs.
+
+## Open questions to mature next
+
+- Concrete acceptance thresholds per category (donation vs listing vs support).
+- CLA / DCO / IP-transfer instrument before any real donation completes (gates
+  the SummaryBot #33 pilot and any future donation; needs legal counsel).
+- Snyk / package-risk evaluation as a standardized pre-review check.
+- Definition of "agentic AI ecosystem" scope.
+
+## Relationship to the charter
+
+The charter (`committee-charter.md`, issue #9) defines mission and organizational
+scope, flowing from the Agentics Foundation profile README. These operating
+principles are the committee's working rules of engagement underneath that
+charter. They are kept short and accrete testable definitions through issues
+rather than swelling into a document.