This document records the design intent and license boundaries for
researcher_agent. It is the authoritative reference when license
edge cases arise.
researcher_agent is a source-available scholarly research copilot
designed for human-in-the-loop use. It assists a human researcher
through the standard academic lifecycle (question → literature →
synthesis → manuscript → review → revision). It is distributed as
a suite of Claude Code skills, slash commands, and sub-agent personas.
- It is not a fully autonomous research system. It will not initiate studies, decide what to argue, or commit to a methodology without human approval.
- It is not a generator of submission-ready manuscripts without human review. Every deliverable is a draft that requires the user to read, evaluate, edit, and approve.
- It is not a replacement for peer review. The "critique" skill (Phase 3) simulates a reviewer's perspective but does not constitute an actual independent peer review.
- It is not a tool for concealing AI involvement. A disclosure command is provided so that AI assistance can be documented in venue-appropriate language.
- Personal academic work: thesis, dissertation, journal articles, conference papers, grant proposals
- Teaching and training: using the framework to demonstrate methodology to students; using sub-agents as examples in research-methods courses
- Non-profit collaboration: research groups, labs, and academic departments sharing the framework internally
- Adaptation for personal use: forking, modifying, and using your own variant in your own research
- Commercial SaaS or hosted services built on the framework
- Consulting, freelance, or paid services that deliver outputs generated primarily by this framework
- Internal use at for-profit organizations where the framework's outputs contribute to revenue-generating work
- Repackaging, rebranding, or reselling the framework as a paid product
- Wrapping the framework in a commercial API
- Using the framework to produce content for clients who pay for it
The CC BY-NC 4.0 license uses "commercial" in the sense of "primarily intended for or directed toward commercial advantage or monetary compensation." Edge cases (e.g., a researcher employed by a for-profit who uses the tool for genuinely non-commercial publication) should be resolved by considering the dominant purpose of the use.
- Submitting AI-assisted output as solely human-authored without any disclosure (an academic integrity issue separate from license)
- Treating critique-skill output as a substitute for actual peer review
- Running the orchestrator end-to-end without engaging with the intermediate deliverables (defeats the human-in-the-loop design)
- Using the framework to produce volume rather than quality
The framework's job is the part that requires patience but not intellectual judgment. Searching multiple databases for relevant sources, formatting citations, checking for consistency across sections, verifying that claimed sources actually say what they are claimed to say — these are time-consuming, error-prone, and not where human cognitive effort produces the most value.
The user's job is the part that requires judgment. Choosing the research question, deciding what evidence counts, interpreting what the data mean, writing the sentences that frame the argument — these are the parts of scholarship that define the work as the user's.
Checkpoints are structural, not stylistic. They are not "would you like me to continue?" friction. They are gates at points where the framework is about to commit to a decision that would be costly to undo (e.g., scope of a systematic review, inclusion/exclusion criteria, final manuscript structure). The user must explicitly approve at these points.
Failure modes are surfaced, not hidden. If the framework cannot locate evidence for a claim, it says so rather than fabricating a citation. If two retrieved sources contradict each other, it records the contradiction rather than picking one silently. If a revision loop reaches the limit, remaining issues are documented as known limitations rather than glossed.
This is original work, written from scratch. The framework references common academic methodologies (PRISMA 2020, Socratic dialogue, peer review protocols, APA 7) which are public-domain scholarly practices. No prompt text, agent specification, template, or example is borrowed or paraphrased from any other research-skills tool.
For uses that fall outside the CC BY-NC 4.0 allowance, contact the copyright holder for commercial licensing terms. Reasonable arrangements can be made for genuine use cases.
Gençer, O. (2026). researcher_agent: An LLM-driven research assistance
framework [Computer software]. CC BY-NC 4.0.