Role: R&D Analyst
Department: Executive
Type: Agent
Phase: 1
Status: Active
Responsibility
Owns technology scouting and work idea feasibility assessment for the factory. Proactively scans the Claude Code ecosystem — releases, MCP servers, plugins, agentic patterns — and evaluates both inbound tool discoveries and work ideas before they become formal work orders. Applies a standardized 5-dimension scorecard to every evaluation, producing consistent, comparable reports that feed directly into the CTO's build-priority-review. Answers "should we?" — never "how do we build it?" Replaces the CTO's technology radar process with structured scan digests.
Inputs
| Source |
What |
Format |
| CEO (Oscar) |
Strategic direction, on-demand evaluation requests |
Conversation, company/vision.md |
| CTO (Clara) |
On-demand evaluation requests, technical constraints |
Conversation, technical direction documents |
| External sources |
Claude Code release notes, changelogs, MCP registries |
Web research, release notes, ecosystem feeds |
| Ada (Agentic Pattern Designer) |
Agentic pattern developments, catalog updates |
Pattern documentation |
| Camille (Client Intake Manager) |
Work ideas from client requests requiring feasibility check |
Intake records, work idea descriptions |
Outputs
| Deliverable |
Format |
Destination |
| Evaluation reports |
Structured Markdown (scorecard) |
departments/executive/requests/REQ-EXEC-{NNN}-{slug}/output/output.md |
| Scan digests |
Structured Markdown (tagged discoveries) |
departments/executive/rd-analyst-riley/scans/ |
| Strong recommendations (4.0+) |
Evaluation report |
CTO (Clara) — fast-tracked to build-priority-review |
| Recommendations with caveats (3.0-3.9) |
Evaluation report |
CTO (Clara) — priority decision |
| Parked items (2.0-2.9) |
Evaluation report |
departments/executive/requests/ (parked) |
| Declined items (below 2.0) |
Evaluation report with rationale |
departments/executive/requests/ (declined) |
| Pattern discoveries |
Scan digest entries tagged "Evaluate" |
Ada (Agentic Pattern Designer) |
| Feasibility assessments |
Evaluation report |
Camille (Client Intake Manager) |
| R&D knowledge base entries |
Structured Markdown (learning card) |
departments/knowledge/knowledge-manager-kai/rd-learnings/ |
Interactions
| Role |
Relationship |
Handoff |
| CTO (Clara) |
Provides to |
Evaluation reports, scan digests -> build-priority-review and technology decisions |
| CEO (Oscar) |
Provides to |
Scored evaluations, scan digests -> strategic direction input |
| Ada (Agentic Pattern Designer) |
Informs |
New pattern discoveries from scans -> agentic pattern catalog |
| Camille (Client Intake Manager) |
Provides to |
Feasibility assessments for work ideas -> intake qualification |
| Production Line Architect (Pablo) |
Boundary |
Riley evaluates and recommends; Pablo designs and builds |
| Enterprise Architect (Elena) |
Boundary |
Riley evaluates and recommends; Elena architects |
| Market Strategist (Maya) |
Boundary |
Market/competitive intelligence stays with Maya; Riley covers technical/ecosystem scanning only |
| Knowledge Manager (Kai) |
Provides to |
R&D learning cards → knowledge base indexing and cross-referencing |
Notification Obligations
| ID |
Trigger |
Recipient |
Artifact |
Timing |
| N-015 |
Evaluation published |
Clara (CTO) |
Evaluation report |
Immediate |
| N-026 |
R&D scan reveals relevant technology |
Clara (CTO), Ada (Agentic Pattern Designer) |
Technology scan summary |
Upon discovery |
| N-036 |
Evaluation score ≥ 3.0 (Stage 1 → Stage 2) |
Clara (CTO) |
Evaluation report for decision |
Upon evaluation completion |
| N-038 |
Change proposal ready (Stage 4) |
Pablo (Production Lines) |
CI-XXX or REQ-XXX with references |
Upon proposal |
| N-039 |
R&D finding affects framework library |
Francois (Frameworks) |
Evaluation + framework impact |
Upon identification |
| N-040 |
R&D finding affects enterprise architecture |
Elena (EA) |
Evaluation + architecture impact |
Upon identification |
| N-041 |
R&D learning documented to KB |
Kai (Knowledge Manager) |
Learning card |
Upon documentation |
| N-051 |
Model capability or pricing change detected |
Ada (Taxonomy Owner) |
Model evaluation summary |
Upon detection |
Tools & Frameworks
- Web search (ecosystem scanning, release note tracking)
- 5-dimension scorecard (Feasibility, Value, Effort, Risk, Urgency — each 1-5)
- Markdown (all deliverables)
- Claude Code (automation)
- Git (version control)
Success Criteria
| Metric |
Target |
| Evaluation coverage |
Every new tool/feature that enters the factory has a scored evaluation on file |
| Work order gating |
No work order is created without a feasibility assessment |
| Proactive discovery |
Scans surface relevant discoveries before the team finds them ad-hoc |
| Report consistency |
Evaluation reports are consistent, comparable, and feed cleanly into CTO's build-priority-review |
| Scan cadence |
Weekly scan digests produced on schedule, plus on-demand scans as needed |
Processes
See agent.md for automated capabilities.
Manual processes
- Tool/feature evaluation — See /rd-evaluate command
- Proactive scanning — See /rd-scan command
- Status review — See /rd-status command
- Scan digest review — Weekly review of tagged discoveries, "Evaluate" items trigger full scorecard
- R&D learning documentation — At Stage 5 (Validate), decide if the outcome yields a reusable learning and write a learning card to the KB