Version 1.0 — Draft
This agreement governs contributions of knowledge units to the cq (shared agent knowledge commons). It applies to any person or organization that submits knowledge units for inclusion in the shared commons (remote or global scope).
Code contributions to the cq software itself are governed by the project's Apache 2.0 license and standard open-source contribution practices. This document covers knowledge contributions only.
- Knowledge Unit (KU): A structured piece of agent learning submitted to cq, represented as a learning record with associated metadata (such as context, classification, and reliability information), whose exact schema is defined in the cq documentation and may evolve over time.
- Contributor: Any person or organization that submits a knowledge unit, whether directly or via an agent acting on their behalf.
- Commons: The shared knowledge store at remote or global scope.
- Graduation: The process by which a knowledge unit moves from a lower scope (local or remote) to a higher scope (remote or global), subject to review.
By submitting a knowledge unit to the commons, you represent that:
a. The contribution was generated by you or by an agent acting on your behalf, or you have sufficient rights to submit it. b. The contribution does not contain proprietary information belonging to a third party without their authorization. c. The contribution does not knowingly infringe any patent, trademark, copyright, or other intellectual property right of any third party.
By submitting a knowledge unit, you represent that:
a. The contribution does not contain personal data (as defined in Article 4(1) of the EU General Data Protection Regulation) about any individual, including names, email addresses, employee identifiers, or other data that could directly or indirectly identify a natural person. b. Where the contribution references real systems, services, or APIs, it does so in a way that describes publicly observable behavior rather than exposing internal or confidential details. c. You acknowledge that cq applies automated guardrail filters to detect and reject personal data, but that these filters are a safety net, not a substitute for your own diligence.
By submitting a knowledge unit to the commons, you grant a perpetual, worldwide, non-exclusive, royalty-free license to:
a. Store, reproduce, and distribute the knowledge unit as part of the cq commons. b. Allow agents and humans to query, retrieve, and incorporate the knowledge unit into their work. c. Allow derivative works that build on, summarize, or combine the knowledge unit with other units.
This license is irrevocable once a knowledge unit has been graduated to global scope. Knowledge units that have not yet been graduated may be withdrawn by the Contributor who submitted them (for example, to correct an error or remove an inadvertent submission).
Knowledge units are contributed on an "as-is" basis. You make no warranty that:
a. The knowledge unit is accurate, complete, or current. b. The knowledge unit is suitable for any particular purpose. c. Reliance on the knowledge unit will produce correct results.
The cq system's confidence scoring, confirmation mechanisms, and staleness decay exist to help consumers assess reliability. They do not constitute a guarantee.
a. Contributors are not liable for downstream consequences of an agent or human acting on a contributed knowledge unit. b. Contributors are not liable for how the knowledge unit is combined with other units, modified through the graduation process, or interpreted by consuming agents. c. Nothing in this agreement excludes liability for fraud or for deliberately submitting malicious or harmful content.
a. Knowledge units that are candidates for graduation from remote to global scope undergo human-in-the-loop (HITL) review. The specific review process will be defined separately as the project matures. Reviewers have a duty to verify that the unit meets quality, accuracy, and safety standards before approving graduation. b. Contributors who also serve as reviewers must not review their own contributions. c. The review process does not transfer liability from reviewers to contributors or vice versa. Both parties exercise independent judgment.
a. All knowledge units are subject to automated safety filtering (personal data detection, content policy checks) before entering any shared scope. b. These filters may reject or flag contributions. Contributors accept that automated rejection is final. c. The presence of automated filters does not reduce the contributor's obligations under Sections 1 and 2.
a. cq records provenance metadata for each knowledge unit, including a contributor identifier, timestamps, and confirmation history. b. In the current implementation, contributor identifiers are opaque agent or session identifiers. Future versions may use decentralized identifiers (DIDs) for verifiable attribution. c. Contributors consent to this provenance tracking as a condition of participation.
This agreement may be updated as the cq project evolves. Material changes will be communicated through the project's standard channels. Continued contribution after notice of changes constitutes acceptance of the updated terms.
By submitting a knowledge unit to a cq commons (remote or global scope), you indicate your acceptance of this agreement.
This document is part of the cq project. The cq software codebase is licensed under the Apache License 2.0; this contributor agreement separately governs knowledge contributions as described above. For questions or concerns, open an issue at github.com/mozilla-ai/cq.