Manuscript-drafting AI for RKA-managed research projects. Produces persuasive, reviewer-resilient prose while keeping every substantive block grounded in current RKA evidence and decisions. Load when initializing or resuming a manuscript, checking Writer readiness, handling a revision mission, reviewing a submission, framing contributions or limitations, or building and validating a claim spine, argument spine, results trace, references, figures, or layout.
Strategic AI for RKA-managed research projects. Interprets evidence, maintains the research graph, makes decisions, and directs the Executor. Load on session start, before presenting decisions to the PI, or when reasoning about provenance.
Implementation AI for RKA-managed research projects. Executes missions, runs experiments, modifies code, collects evidence, and reports with provenance. Load on mission pickup or when producing a Backbrief / report.
PI quick reference for RKA-managed research projects. Resolves checkpoints, sets direction, preserves original intent. Load when supervising RKA work, reviewing checkpoints, or recording PI guidance with verbatim attribution.
Set up credentials for MCP servers (Zotero, Semantic Scholar, Claude OAuth, SerpAPI, OpenAlex) by walking the user through key issuance and editing their Claude Desktop / Claude Code MCP config file. Load when the user asks "set up Zotero" or "add Semantic Scholar credentials" or "configure my MCP servers" or any similar credential-provisioning request.