dd-gap-analysis

name	dd-gap-analysis
description	Run a structured gap analysis on any design document or project folder. Use when asked to check what's missing, find gaps, or review completeness of a design.
argument-hint	[project folder or name]
version	0.1.0
allowed-tools	["Read","Glob","Grep"]

Gap Analysis

A standalone skill that analyzes design documents for gaps, missing coverage, and open work. This skill has no dependency on the Daydream Dictation process — it works on any set of documents at any time. A user might invoke it during a Phase 2 session, run it cold on a project they haven't touched in weeks, or use it as the first step when picking up an unfamiliar document.

How to Execute

1. Resolve the target project:

   • If $ARGUMENTS names a project, find its folder
   • If $ARGUMENTS is a path, use it directly
   • If empty, read dd-current-dictation-project for the active project
   • If still nothing, ask the user which project to analyze

2. Read the project's documents:

   • Daydream-<Slug>.md — the main design document
   • TODO-<Slug>.md — the to-do list
   • Prompts-<Slug>.md — tail of the prompts log (last 20–30 entries for recent context)

   Not all project folders follow this exact naming. If the folder uses different filenames, look for a root document that serves as the main design doc (e.g., Overview-*.md) and use whatever documentation is present. Adapt to the folder's structure rather than skipping analysis because filenames don't match.

3. Answer each of the seven questions below. For each one, give concrete findings — not generic advice. Quote specific sections, name specific gaps, reference specific TODOs.

4. After answering all seven, write a summary of actionable items and offer to address them.

• Standalone output. The gap analysis produces a self-contained report. Do not modify any documents — only read, report and offer to help. The user decides what to act on.

---

The Seven Questions

1. What are we missing from the design?

Are there systems, features, or areas that have been mentioned or implied but not yet documented? Are there sections that feel thin or underspecified? Look for things the user talked about in prompts that never made it into the document.

2. What to-do items are still open?

Review TODO-<Slug>.md. Which items remain unresolved? Scan the Prompts log and the design document for to-do items that were mentioned but never added to the TODO file. Report any discrepancies.

3. What are opportunities for improvement?

Beyond what's missing — where does the existing design feel unclear, inconsistent, or underdeveloped? What could be strengthened? Look for contradictions between sections, vague language that would block implementation, or areas where the user changed their mind but the document wasn't fully updated.

4. What will the implementing agent need to know?

Identify anything that an agent picking up this design document would find ambiguous or underspecified. What decisions would they have to make on their own because the document doesn't answer them? If the project has a DecisionTrace-<Slug>.md, check whether it covers the technical decisions needed to implement the design. Does the document call out open decisions? Does the user want to make them now? If no Decision Trace exists and the project has enough technical complexity to warrant one, note that.

5. Are there missing control definitions?

For any interactive or playable system: are the player controls fully defined? Are all inputs accounted for across all relevant states? Skip this question if the project is not interactive.

6. Are there untranslated strings?

If the project has a StringTable-<Slug>.md or other localization files, check for completeness: are there strings missing translations for any supported language? Are there strings referenced in the design doc that don't have string table entries? If gaps exist, ask: should translation happen now, or is it deferred? Skip this question if the project has no localization.

7. Are there any other gaps?

Catch-all for anything not covered above — missing asset lists, undefined edge cases, unresolved working names, placeholder items ([Item N — not yet documented]), or anything else that would block a complete implementation.

---

Output Format

For each question, respond with:

• Finding — what you found (or "No gaps identified" if clean)
• Severity — critical (blocks implementation), moderate (should be addressed), minor (nice to have)
• Suggestion — what to do about it

End with a numbered list of all actionable items sorted by severity, and ask the user which ones to tackle.

Goal

The gap analysis is a signal that the design is complete enough to hand off. By the time all seven questions are answered satisfactorily, you should be confident that an implementing agent has everything it needs, and that the user has touched every part of the design before moving on. If significant gaps remain, say so clearly. If the design looks solid, tell the user it's ready.