Planning With Files

OpenClaw Integration: This skill is invoked by the Dojo Genesis plugin via /dojo run planning-with-files. The agent receives project context automatically via the before_agent_start hook. Use dojo_get_context for full state, dojo_save_artifact to persist outputs, and dojo_update_state to record phase transitions and decisions.

Planning with Files (Meta-Skill)

Version: 1.0
Created: 2026-02-08
Author: Manus AI
Purpose: To provide a unified interface for all file-based planning workflows while maintaining the specialized quality of each mode.

I. The Philosophy: One Entry, Many Paths

This skill acts as an intelligent router, analyzing the user's request and uploaded files to determine the most appropriate specialized workflow. It embodies the principle of intent-based routing, ensuring that the right tool is used for the right job without requiring the user to know the names of the individual skills.

II. When to Use This Skill

• When you need to create a plan from uploaded files (specs, docs, code, research).
• When you want to write a specification for a new feature or refactoring task.
• When you need to synthesize multiple research files into actionable insights.
• When you have files and a planning goal, but you're not sure which skill to use.

III. The Workflow

This is a 3-step workflow for routing and executing file-based planning tasks.

Step 1: File and Intent Analysis

Goal: Determine which mode is most appropriate for the user's request.

1. Analyze Files: Read all uploaded files to understand their type, content, and quantity.
2. Analyze Intent: Analyze the user's request for keywords and intent signals.
3. Apply Routing Logic: Use the routing table below to select the appropriate mode.

Routing Table:

Step 2: Mode Routing and Execution

Goal: Hand off the task to the selected specialized mode and monitor execution.

1. Invoke Skill: Load the selected skill's workflow (context-ingestion, specification-driven-development, or research-synthesis).
2. Pass Context: Pass the original files and request to the selected skill.
3. Monitor Execution: Track the progress of the specialized skill.

Step 3: Delivery and Feedback

Goal: Deliver the results to the user and gather feedback for routing improvement.

1. Present Output: Deliver the output file(s) from the specialized skill.
2. Explain Routing: Briefly explain why the mode was selected (e.g., "I used the research synthesis engine because you uploaded multiple research papers.").
3. Gather Feedback: Ask if the selected mode was appropriate and log feedback for future routing improvements.

IV. Best Practices

• Trust the Router: The routing logic is designed to select the best mode for the job. Let the skill do its work.
• Provide Clear Intent: The more specific your request, the better the routing will be (e.g., "synthesize these papers" is better than "look at these files").
• User Override: If the skill selects the wrong mode, you can always explicitly invoke the correct skill.

V. Quality Checklist

• Have you uploaded all relevant files for the planning task?
• Have you provided a clear request with intent keywords (plan, spec, synthesize)?
• Have you reviewed the output to ensure it matches your expectations?
• Have you provided feedback on the routing decision if it was incorrect?

VI. Common Pitfalls

Underspecified Intent

Problem: Vague requests like "look at these files" force the router to guess. Guessing defaults to context-ingestion, which may not be what you wanted.

Solution: Include an intent keyword: "plan from these files," "synthesize these papers," or "spec this feature." One word changes the routing.

Mixing File Types Without Context

Problem: Uploading a spec, three research papers, and a code file together. The router sees conflicting signals and can't determine which workflow you need.

Solution: Either group files by purpose (research together, specs together) or state your intent explicitly. The router handles ambiguity, but clarity is faster.

Skipping the Routing Explanation

Problem: Jumping straight to the output without telling the user which mode was selected. If the mode was wrong, the user won't know until the output is off.

Solution: Always state which mode was selected and why before delivering results. One sentence is enough: "I routed this to research-synthesis because you uploaded 4 research PDFs."

VII. Example

Context: A user uploads two files — a backend architecture doc and a product brief — and says "help me plan the frontend for this."

Routing decision: The word "plan" + 2 files → context-ingestion. But "frontend" + "backend architecture doc" is a stronger signal for frontend-from-backend. The router selects frontend-from-backend based on the file type override.

Outcome: A frontend specification grounded in the actual backend endpoints, data models, and auth patterns from the uploaded architecture doc.

VIII. Related Skills

• context-ingestion — Default routing target for general planning from files
• research-synthesis — Target for multi-paper synthesis
• release-specification — Target when files indicate a spec-writing workflow
• frontend-from-backend — Target when backend architecture docs are present with frontend intent

OpenClaw Tool Integration

When running inside the Dojo Genesis plugin:

1. Start by calling dojo_get_context to retrieve full project state, history, and artifacts
2. During the skill, follow the workflow steps documented above
3. Save outputs using dojo_save_artifact with the artifacts output directory
4. Update project state by calling dojo_update_state to record skill completion and any phase transitions