feat: add research prompt template library (skills/research/)

Create 6 structured research prompt templates with YAML frontmatter,
cascade_tier hints, and {slot} placeholders:

- tool_evaluation.md — discover and compare tools in a domain
- architecture_spike.md — investigate system integration approaches
- game_analysis.md — evaluate games for AI agent playability
- integration_guide.md — step-by-step tool integration with code
- state_of_art.md — landscape survey of a field at a point in time
- competitive_scan.md — compare a project against alternatives

Each template has valid YAML frontmatter (name, type, query count,
output length, cascade_tier) and produces well-structured prompts
when slots are filled.

Fixes #974

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

skills/research/architecture_spike.md

@@ -0,0 +1,67 @@
+---
+name: Architecture Spike
+type: research
+typical_query_count: 2-4
+expected_output_length: 600-1200 words
+cascade_tier: groq_preferred
+description: >
+  Investigate how to connect two systems or components. Produces an integration
+  architecture with sequence diagram, key decisions, and a proof-of-concept outline.
+---
+
+# Architecture Spike: Connect {system_a} to {system_b}
+
+## Context
+
+We need to integrate **{system_a}** with **{system_b}** in the context of
+**{project_context}**. This spike answers: what is the best way to wire them
+together, and what are the trade-offs?
+
+## Constraints
+
+- Prefer approaches that avoid adding new infrastructure dependencies.
+- The integration should be **{sync_or_async}** (synchronous / asynchronous).
+- Must work within: {environment_constraints}.
+
+## Research Steps
+
+1. Identify the APIs / protocols exposed by both systems.
+2. List all known integration patterns (direct API, message queue, webhook, SDK, etc.).
+3. Evaluate each pattern for complexity, reliability, and latency.
+4. Select the recommended approach and outline a proof-of-concept.
+
+## Output Format
+
+### Integration Options
+
+| Pattern | Complexity | Reliability | Latency | Notes |
+|---------|-----------|-------------|---------|-------|
+| ...     | ...       | ...         | ...     | ...   |
+
+### Recommended Approach
+
+**Pattern:** {pattern_name}
+
+**Why:** One paragraph explaining the choice.
+
+### Sequence Diagram
+
+```
+{system_a} -> {middleware} -> {system_b}
+```
+
+Describe the data flow step by step:
+
+1. {system_a} does X...
+2. {middleware} transforms / routes...
+3. {system_b} receives Y...
+
+### Proof-of-Concept Outline
+
+- Files to create or modify
+- Key libraries / dependencies needed
+- Estimated effort: {effort_estimate}
+
+### Open Questions
+
+Bullet list of decisions that need human input before proceeding.

skills/research/competitive_scan.md

@@ -0,0 +1,74 @@
+---
+name: Competitive Scan
+type: research
+typical_query_count: 3-5
+expected_output_length: 800-1500 words
+cascade_tier: groq_preferred
+description: >
+  Compare a project against its alternatives. Produces a feature matrix,
+  strengths/weaknesses analysis, and positioning summary.
+---
+
+# Competitive Scan: {project} vs Alternatives
+
+## Context
+
+Compare **{project}** against **{alternatives}** (comma-separated list of
+competitors). The goal is to understand where {project} stands and identify
+differentiation opportunities.
+
+## Constraints
+
+- Comparison date: {date}.
+- Focus areas: {focus_areas} (e.g., features, pricing, community, performance).
+- Perspective: {perspective} (user, developer, business).
+
+## Research Steps
+
+1. Gather key facts about {project} (features, pricing, community size, release cadence).
+2. Gather the same data for each alternative in {alternatives}.
+3. Build a feature comparison matrix.
+4. Identify strengths and weaknesses for each entry.
+5. Summarize positioning and recommend next steps.
+
+## Output Format
+
+### Overview
+
+One paragraph: what space does {project} compete in, and who are the main players?
+
+### Feature Matrix
+
+| Feature / Attribute | {project} | {alt_1} | {alt_2} | {alt_3} |
+|--------------------|-----------|---------|---------|---------|
+| {feature_1}        | ...       | ...     | ...     | ...     |
+| {feature_2}        | ...       | ...     | ...     | ...     |
+| Pricing            | ...       | ...     | ...     | ...     |
+| License            | ...       | ...     | ...     | ...     |
+| Community Size     | ...       | ...     | ...     | ...     |
+| Last Major Release | ...       | ...     | ...     | ...     |
+
+### Strengths & Weaknesses
+
+#### {project}
+- **Strengths:** ...
+- **Weaknesses:** ...
+
+#### {alt_1}
+- **Strengths:** ...
+- **Weaknesses:** ...
+
+_(Repeat for each alternative)_
+
+### Positioning Map
+
+Describe where each project sits along the key dimensions (e.g., simplicity
+vs power, free vs paid, niche vs general).
+
+### Recommendations
+
+Bullet list of actions based on the competitive landscape:
+
+- **Differentiate on:** {differentiator}
+- **Watch out for:** {threat}
+- **Consider adopting from {alt}:** {feature_or_approach}

skills/research/game_analysis.md

@@ -0,0 +1,68 @@
+---
+name: Game Analysis
+type: research
+typical_query_count: 2-3
+expected_output_length: 600-1000 words
+cascade_tier: local_ok
+description: >
+  Evaluate a game for AI agent playability. Assesses API availability,
+  observation/action spaces, and existing bot ecosystems.
+---
+
+# Game Analysis: {game}
+
+## Context
+
+Evaluate **{game}** to determine whether an AI agent can play it effectively.
+Focus on programmatic access, observation space, action space, and existing
+bot/AI ecosystems.
+
+## Constraints
+
+- Platform: {platform} (PC, console, mobile, browser).
+- Agent type: {agent_type} (reinforcement learning, rule-based, LLM-driven, hybrid).
+- Budget for API/licenses: {budget}.
+
+## Research Steps
+
+1. Identify official APIs, modding support, or programmatic access methods for {game}.
+2. Characterize the observation space (screen pixels, game state JSON, memory reading, etc.).
+3. Characterize the action space (keyboard/mouse, API calls, controller inputs).
+4. Survey existing bots, AI projects, or research papers for {game}.
+5. Assess feasibility and difficulty for the target agent type.
+
+## Output Format
+
+### Game Profile
+
+| Property          | Value                  |
+|-------------------|------------------------|
+| Game              | {game}                 |
+| Genre             | {genre}                |
+| Platform          | {platform}             |
+| API Available     | Yes / No / Partial     |
+| Mod Support       | Yes / No / Limited     |
+| Existing AI Work  | Extensive / Some / None|
+
+### Observation Space
+
+Describe what data the agent can access and how (API, screen capture, memory hooks, etc.).
+
+### Action Space
+
+Describe how the agent can interact with the game (input methods, timing constraints, etc.).
+
+### Existing Ecosystem
+
+List known bots, frameworks, research papers, or communities working on AI for {game}.
+
+### Feasibility Assessment
+
+- **Difficulty:** Easy / Medium / Hard / Impractical
+- **Best approach:** {recommended_agent_type}
+- **Key challenges:** Bullet list
+- **Estimated time to MVP:** {time_estimate}
+
+### Recommendation
+
+One paragraph: should we proceed, and if so, what is the first step?

skills/research/integration_guide.md

@@ -0,0 +1,79 @@
+---
+name: Integration Guide
+type: research
+typical_query_count: 3-5
+expected_output_length: 1000-2000 words
+cascade_tier: groq_preferred
+description: >
+  Step-by-step guide to wire a specific tool into an existing stack,
+  complete with code samples, configuration, and testing steps.
+---
+
+# Integration Guide: Wire {tool} into {stack}
+
+## Context
+
+Integrate **{tool}** into our **{stack}** stack. The goal is to
+**{integration_goal}** (e.g., "add vector search to the dashboard",
+"send notifications via Telegram").
+
+## Constraints
+
+- Must follow existing project conventions (see CLAUDE.md).
+- No new cloud AI dependencies unless explicitly approved.
+- Environment config via `pydantic-settings` / `config.py`.
+
+## Research Steps
+
+1. Review {tool}'s official documentation for installation and setup.
+2. Identify the minimal dependency set required.
+3. Map {tool}'s API to our existing patterns (singletons, graceful degradation).
+4. Write integration code with proper error handling.
+5. Define configuration variables and their defaults.
+
+## Output Format
+
+### Prerequisites
+
+- Dependencies to install (with versions)
+- External services or accounts required
+- Environment variables to configure
+
+### Configuration
+
+```python
+# In config.py — add these fields to Settings:
+{config_fields}
+```
+
+### Implementation
+
+```python
+# {file_path}
+{implementation_code}
+```
+
+### Graceful Degradation
+
+Describe how the integration behaves when {tool} is unavailable:
+
+| Scenario              | Behavior           | Log Level |
+|-----------------------|--------------------|-----------|
+| {tool} not installed  | {fallback}         | WARNING   |
+| {tool} unreachable    | {fallback}         | WARNING   |
+| Invalid credentials   | {fallback}         | ERROR     |
+
+### Testing
+
+```python
+# tests/unit/test_{tool_snake}.py
+{test_code}
+```
+
+### Verification Checklist
+
+- [ ] Dependency added to pyproject.toml
+- [ ] Config fields added with sensible defaults
+- [ ] Graceful degradation tested (service down)
+- [ ] Unit tests pass (`tox -e unit`)
+- [ ] No new linting errors (`tox -e lint`)

skills/research/state_of_art.md

@@ -0,0 +1,67 @@
+---
+name: State of the Art
+type: research
+typical_query_count: 4-6
+expected_output_length: 1000-2000 words
+cascade_tier: groq_preferred
+description: >
+  Comprehensive survey of what currently exists in a given field or domain.
+  Produces a structured landscape overview with key players, trends, and gaps.
+---
+
+# State of the Art: {field} (as of {date})
+
+## Context
+
+Survey the current landscape of **{field}**. Identify key players, recent
+developments, dominant approaches, and notable gaps. This is a point-in-time
+snapshot intended to inform decision-making.
+
+## Constraints
+
+- Focus on developments from the last {timeframe} (e.g., 12 months, 2 years).
+- Prioritize {priority} (open-source, commercial, academic, or all).
+- Target audience: {audience} (technical team, leadership, general).
+
+## Research Steps
+
+1. Identify the major categories or sub-domains within {field}.
+2. For each category, list the leading projects, companies, or research groups.
+3. Note recent milestones, releases, or breakthroughs.
+4. Identify emerging trends and directions.
+5. Highlight gaps — things that don't exist yet but should.
+
+## Output Format
+
+### Executive Summary
+
+Two to three sentences: what is the state of {field} right now?
+
+### Landscape Map
+
+| Category       | Key Players              | Maturity    | Trend       |
+|---------------|--------------------------|-------------|-------------|
+| {category_1}  | {player_a}, {player_b}   | Early / GA  | Growing / Stable / Declining |
+| {category_2}  | {player_c}, {player_d}   | Early / GA  | Growing / Stable / Declining |
+
+### Recent Milestones
+
+Chronological list of notable events in the last {timeframe}:
+
+- **{date_1}:** {event_description}
+- **{date_2}:** {event_description}
+
+### Trends
+
+Numbered list of the top 3-5 trends shaping {field}:
+
+1. **{trend_name}** — {one-line description}
+2. **{trend_name}** — {one-line description}
+
+### Gaps & Opportunities
+
+Bullet list of things that are missing, underdeveloped, or ripe for innovation.
+
+### Implications for Us
+
+One paragraph: what does this mean for our project? What should we do next?

skills/research/tool_evaluation.md

@@ -0,0 +1,52 @@
+---
+name: Tool Evaluation
+type: research
+typical_query_count: 3-5
+expected_output_length: 800-1500 words
+cascade_tier: groq_preferred
+description: >
+  Discover and evaluate all shipping tools/libraries/services in a given domain.
+  Produces a ranked comparison table with pros, cons, and recommendation.
+---
+
+# Tool Evaluation: {domain}
+
+## Context
+
+You are researching tools, libraries, and services for **{domain}**.
+The goal is to find everything that is currently shipping (not vaporware)
+and produce a structured comparison.
+
+## Constraints
+
+- Only include tools that have public releases or hosted services available today.
+- If a tool is in beta/preview, note that clearly.
+- Focus on {focus_criteria} when evaluating (e.g., cost, ease of integration, community size).
+
+## Research Steps
+
+1. Identify all actively-maintained tools in the **{domain}** space.
+2. For each tool, gather: name, URL, license/pricing, last release date, language/platform.
+3. Evaluate each tool against the focus criteria.
+4. Rank by overall fit for the use case: **{use_case}**.
+
+## Output Format
+
+### Summary
+
+One paragraph: what the landscape looks like and the top recommendation.
+
+### Comparison Table
+
+| Tool | License / Price | Last Release | Language | {focus_criteria} Score | Notes |
+|------|----------------|--------------|----------|----------------------|-------|
+| ...  | ...            | ...          | ...      | ...                  | ...   |
+
+### Top Pick
+
+- **Recommended:** {tool_name} — {one-line reason}
+- **Runner-up:** {tool_name} — {one-line reason}
+
+### Risks & Gaps
+
+Bullet list of things to watch out for (missing features, vendor lock-in, etc.).