# Plan Research-Backed Software Projects

Complex projects fail when teams plan from assumptions and miss known risks. This skill guides evidence collection, task planning, and pre-mortem review before implementation.

## Install

```bash
npx skillstore add dnyoussef/research-driven-planning
```

## Metadata

- Status: approved
- Slug: dnyoussef-research-driven-planning
- Skillstore revision: r1
- Version status: missing
- Tree hash: aa1a97df5b08f68afb6cd785c126af91a730d2ddba2d86cecbc6e2b56291becd
- Author: DNYoussef
- GitHub username: DNYoussef
- License: MIT
- Repository: https://github.com/DNYoussef/ai-chrome-extension/tree/main/.claude/skills/research-driven-planning
- Ref: c579efd02184f1128d6741d16e6c8d951d7d4208
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, filesystem
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/dnyoussef-research-driven-planning
- Manifest: https://skillstore.pages.dev/api/skills/dnyoussef-research-driven-planning/manifest

## Capabilities

- Guides creation of SPEC.md with requirements, constraints, and success criteria.
- Defines a six-agent research workflow for evidence gathering and technology comparison.
- Transforms specifications into an enhanced implementation plan with risks and dependencies.
- Runs iterative pre-mortem review with consensus and failure-confidence checks.
- Packages planning outputs for implementation handoff through files and memory namespaces.

## Use Cases

- Clarify Feature Scope: Turn a broad product idea into requirements, constraints, success criteria, and explicit exclusions.
- Choose a Technical Approach: Compare libraries, patterns, and platform choices using evidence before committing to implementation.
- Review Build Risk Early: Identify failure modes, mitigation strategies, and handoff risks before a development swarm starts.

## Prompt Templates

### Start a Planning Pass

```
Use research-driven-planning for this feature: [feature]. Capture requirements, constraints, success criteria, and out-of-scope items before research.
```

### Compare Implementation Options

```
Use research-driven-planning to compare approaches for [problem]. Include evidence sources, security tradeoffs, reliability risks, and a recommended path.
```

### Run a Pre-Mortem

```
Use research-driven-planning to run a pre-mortem for this plan: [plan summary]. Identify likely failures, root causes, mitigations, and acceptance checks.
```

### Prepare Implementation Handoff

```
Use research-driven-planning to create a handoff package for Loop 2. Include validated requirements, research-backed tasks, risk mitigations, and readiness criteria.
```

## Limitations

- Requires clear project goals, constraints, and success criteria from the user.
- Depends on Claude Flow commands and agent workflows that may not exist in every environment.
- Takes more time than direct implementation and may be too heavy for small fixes.
- Confidence scores depend on the quality and freshness of collected evidence.

## Best Practices

- Provide constraints, current architecture, and success metrics before starting research.
- Review evidence quality and source freshness before accepting any recommendation.
- Use the pre-mortem output to change the plan, not only to document risks.

## Anti Patterns

- Using this workflow for urgent hotfixes or small edits that do not need research.
- Treating consensus scores as proof when evidence sources are weak or outdated.
- Handing off a plan before unresolved critical risks have owners and mitigations.

## Security Audit

- Audited at: 2026-07-05T12:39:48.179\+00:00
- Summary: Most static alerts are false positives caused by Markdown fences, inline code, JavaScript templates, and fixed project artifact paths. The real risks are external command execution through npx claude-flow@alpha and command substitutions that persist project contents into Claude Flow memory. No evidence found for prompt injection, credential theft, destructive file access, or malicious reconnaissance instructions.

## Stats

- Views: 309
- Downloads: 11
- Favorites: 0
- Popularity score: 0