From plan-workflow
This skill should be used when the user asks to "research a feature", "create a feature spec", "analyze external APIs for a feature", "plan feature research", "generate a feature-spec.md", or mentions needing comprehensive research before implementing a new feature. Also triggered by the /feature-research command. Creates feature-spec.md ready for plan-workflow.
How this skill is triggered — by the user, by Claude, or both
Slash command
/plan-workflow:feature-researchThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Deep research skill for application features. Goes beyond codebase analysis to research external APIs, business logic, technical specifications, UX considerations, and generate actionable recommendations.
Deep research skill for application features. Goes beyond codebase analysis to research external APIs, business logic, technical specifications, UX considerations, and generate actionable recommendations.
feature-research --> plan-workflow --> implement-plan
(this skill) (Step 2) (Step 3)
Creates: Uses: Executes:
feature-spec.md feature-spec.md parallel-plan.md
Key Distinction:
shared-context: Codebase-focused research (files, patterns, tables, internal docs)feature-research: Comprehensive research (external APIs, business logic, UX, technical specs, recommendations)Target: $ARGUMENTS
Parse arguments:
docs/plans/If no feature name provided, abort with usage instructions:
Usage: /feature-research [feature-name] [--description "..."] [--dry-run]
Examples:
/feature-research plex-integration --description "Advanced Plex media library integration with filters and playlists"
/feature-research payment-system --description "Stripe payment integration for subscriptions"
/feature-research user-dashboard --dry-run
Extract from $ARGUMENTS:
Validate the feature name: must be kebab-case, no special characters except hyphens.
test -d "docs/plans/[feature-name]" && echo "exists" || mkdir -p "docs/plans/[feature-name]"
List any existing files. If feature-spec.md exists, read it and ask user if they want to regenerate or enhance.
If --dry-run is present, display the execution plan (agents, output files, directory) and STOP without creating files or deploying agents.
Read the research prompts template:
cat ${CLAUDE_PLUGIN_ROOT}/skills/feature-research/templates/research-agents.md
CRITICAL: Deploy all 5 agents in a SINGLE message with MULTIPLE Task tool calls.
| Agent | Subagent Type | Output File | Focus |
|---|---|---|---|
| External API Researcher | research-specialist | research-external.md | External APIs, libraries, documentation, integration patterns |
| Business Logic Analyzer | codebase-research-analyst | research-business.md | Requirements, user stories, business rules, domain logic |
| Technical Spec Designer | codebase-research-analyst | research-technical.md | Architecture, data models, API design, system constraints |
| UX Researcher | research-specialist | research-ux.md | User experience, workflows, best practices, accessibility |
| Recommendations Agent | codebase-research-analyst | research-recommendations.md | Ideas, improvements, related features, risks |
Use the prompts from research-agents.md with variables substituted:
{{FEATURE_NAME}} - The feature directory name{{FEATURE_DIR}} - Full output directory path (docs/plans/[feature-name]){{FEATURE_DESCRIPTION}} - The description provided (or feature name if none)After agents complete, validate all research-*.md files:
${CLAUDE_PLUGIN_ROOT}/skills/feature-research/scripts/validate-research.sh docs/plans/[feature-name]
If validation fails: identify failed workstream files, send targeted corrective follow-up only to owning agents, wait for corrected outputs, rerun validation until pass.
Read all 5 research files from docs/plans/[feature-name]/.
cat ${CLAUDE_PLUGIN_ROOT}/skills/feature-research/templates/spec-structure.md
Create docs/plans/[feature-name]/feature-spec.md following the template.
Key Consolidation Principles:
${CLAUDE_PLUGIN_ROOT}/skills/feature-research/scripts/validate-spec.sh docs/plans/[feature-name]/feature-spec.md
Fix any issues reported (missing sections, empty content, formatting).
Provide completion summary with: feature name, description, files created, research summary counts, key findings, decisions needed, and next steps (review spec, proceed to /plan-workflow, or add requirements).
Each research file should: focus on its specific domain, include concrete examples and links, identify gaps and uncertainties, provide actionable recommendations, and avoid duplication with other research files.
feature-spec.md before validator passnpx claudepluginhub yandy-r/claude-pluginsDevelops feature ideas into structured specifications via guided conversation with architectural analysis, UX input, and decomposition for large features.
Generates structured feature specifications with demoable units, functional requirements, and proof artifacts. Use when defining what to build before writing code.
Plans single features, endpoints, or infrastructure via Socratic conversation to uncover requirements, edge cases, failure modes, and constraints; outputs markdown feature plan artifact before coding.