| name | agentic-coding-product-research |
| description | Research AI coding assistant products, audiences, user stories, social proof, hacks, and unmet needs for apps like Cursor, Claude Code, Codex, Warp Code, Devin, Windsurf, GitHub Copilot, Cline, Aider, and homegrown agent stacks. Use when designing or positioning agentic developer tools, Port Daddy Agent Harbor, swarm consoles, background coding agents, or competitive/product research for AI software engineering. NOT for implementing the coding agent runtime, generic market research, or UX mockups without developer workflow evidence. |
| license | Apache-2.0 |
| allowed-tools | Read,Write,Edit,Bash,Grep,Glob,WebSearch,WebFetch |
| metadata | {"category":"Agent & Orchestration","tags":["agentic-coding","product-research","developer-tools","ai-coding-assistants","port-daddy"],"provenance":{"kind":"first-party","owners":["port-daddy"]},"pairs-with":[{"skill":"agentic-coding-ux-designer","reason":"Turns research findings into low-friction product flows."},{"skill":"swarm-invocation-designer","reason":"Converts unmet coordination needs into multi-agent invocation patterns."},{"skill":"skill-architect","reason":"Keeps research reusable through progressive disclosure and evals."}],"io-contract":{"kind":"deliverable","consumes":["[Truncated]","[Truncated]"],"produces":["[Truncated]","[Truncated]","[Truncated]"]}} |
Agentic Coding Product Research
Build evidence-backed product intelligence for AI coding assistant and software-agent products.
Use This For
- Mapping audiences, jobs, anxieties, and "I keep coming back" moments for agentic coding tools.
- Comparing Cursor, Claude Code, Codex, Warp Code, Devin, Windsurf/Cascade, Copilot, Cline, Aider, OpenHands, and homegrown stacks.
- Finding unmet needs Port Daddy can own: coordination, durable transcripts, review proof, spend control, sandboxing, swarm visibility, and operator trust.
- Turning tech press, official docs, research papers, GitHub traces, Reddit/HN discourse, and failed local dogfood into user stories.
Do Not Use This For
- Implementing a terminal/editor/runtime agent loop directly.
- UI composition without first turning findings into workflow evidence.
- Treating benchmarks as product truth without user stories and review failure modes.
Process
flowchart TD
A[Collect current evidence] --> B[Segment audiences]
B --> C[Extract jobs, pains, craves]
C --> D[Map product mechanics]
D --> E[Find hacks and homegrown tools]
E --> F[Score Port Daddy opportunities]
F --> G[Emit story matrix and unmet-needs brief]
- Collect sources across four lanes: official docs, tech press, academic/benchmark work, and social/homegrown workflows.
- Segment audiences by workflow pressure, not job title alone: solo founder, staff engineer, enterprise admin, maintainer, agent power user, non-developer builder.
- Extract user stories in the format
As <audience>, I want <agentic capability>, so I can <workflow outcome without hidden risk>.
- Identify product mechanics users praise or crave: context attachment, plan/apply/review, checkpoints, worktrees, visual diff, background agents, model choice, and mobile/cloud handoff.
- Identify negative demand: surprise spend, invisible state, stale context, agent collisions, weak rollback, AI support hallucinations, unsafe tool execution, and review burden.
- Translate evidence into Port Daddy opportunities with a proof requirement for each opportunity.
Output Contract
Produce:
audiences: array of audience profiles with jobs, pains, craves, trust thresholds, and comeback triggers.
user_stories: array of stories with evidence sources and Port Daddy implications.
opportunities: ranked Port Daddy product opportunities with proof artifacts required.
risks: skeptical caveats, source limits, and claims needing live verification.
Use scripts/story_matrix.mjs to validate and derive a JSON story matrix from a source manifest.
Anti-Patterns
Benchmark Theater
Novice: "The product with the highest SWE-bench number wins."
Expert: Benchmarks predict only part of adoption. Users come back when the tool reduces start friction, preserves context, proves its work, and makes mistakes recoverable.
Detection: Research omits user stories, review workflows, or rollback mechanics.
Vibe Without Receipts
Novice: "People love it on social, so ship the same chat box."
Expert: Social praise usually compresses a full loop: prompt, context, agent action, visible progress, diff review, tests, PR, and undo. Capture the loop, not the applause.
Detection: No source links, no artifact requirement, no failure case.
Tool-Only Framing
Novice: "Port Daddy should be another coding assistant."
Expert: Port Daddy's wedge is the coordination control plane around assistants: identity, claims, transcripts, spend, sandboxing, review proof, and multi-agent orchestration.
Detection: Proposed feature competes on code generation alone rather than operator control and durable evidence.
References
| File | Load When |
|---|
references/source-map.md | Need current product/source landscape and research citations. |
references/audience-stories.md | Need audience segmentation, user stories, and Port Daddy opportunities. |
examples/expected-output.md | Need the shape of a finished research brief. |
templates/output-template.md | Need a reusable brief template. |
schemas/source-manifest.schema.json | Need to validate research inputs. |
scripts/story_matrix.mjs | Need deterministic user-story matrix generation. |
agents/openai.yaml | Need a subagent descriptor for delegated product research. |
Skill Bundle Index
Every file in this skill, and when to open it. Auto-generated by the repo skill-architect indexer.
root
CHANGELOG.md — Agentic Coding Product Research — Changelog — - Initial skill creation - Core process defined - Reference files added
README.md — Agentic Coding Product Research — Evidence-backed product research for AI coding assistant and software-agent tools.
agents/
examples/
examples/expected-output.md — Example Output: Agentic Coding Product Research — Research question: what should Port Daddy build around AI coding assistants after reviewing Cursor, Claude Code, Codex, Warp Code, Devin, Wi
references/
references/audience-stories.md — Audience Stories And Port Daddy Opportunities — Use this when turning research into product requirements.
references/source-map.md — Source Map: AI Coding Assistant Product Landscape — Use this when grounding product claims in current market evidence.
schemas/
scripts/
templates/
templates/output-template.md — Agentic Coding Product Research Brief — [One sentence naming the product decision, audience, or Port Daddy opportunity.] | Source | Kind | Current As Of | Claim Used | | --- | ---