Video summary

Up & Running with GitHub Spec Kit #7 - The /analyze Command

Main summary

Key takeaways

Technology

What the video covers (GitHub Spec Kit #7 — /analyze command)

After running a /tasks command to create a roadmap of coding agent actions, the speaker introduces an optional /analyze command.

Purpose of /analyze

The /analyze command performs a read-only review of all artifacts created so far (e.g., constitution, spec file, plan files, etc.) to catch:

  • Inconsistencies
  • Conflicts / contradictions (especially against the constitution)
  • Ambiguities
  • Coverage gaps (requirements missing implementation tasks)
  • Duplications (mentioned as part of the analysis goals)

Under the hood: the analyze prompt behavior

The analyze prompt takes user input as an argument in general, but for /analyze, the speaker suggests no additional guidance is typically needed.

The prompt explicitly instructs the agent to:

  • Identify issues such as inconsistencies, duplications, ambiguities
  • Treat the operation as strictly read-only (no code changes)
  • Compare artifacts against the constitution and flag conflicts

It effectively:

  1. Locates the feature directory
  2. Gathers documents/artifacts in that directory
  3. Runs analysis steps and reports results in a table in chat

Demo results: what /analyze outputs

The command generates a large report, including:

Executive summary

  • Example finding: “No critical constitutional violations found”
  • Notes on strong alignment with constitutional principles and coverage of requirements

Tables of issues

Each issue entry includes:

  • ID
  • Category (e.g., ambiguity / inconsistency / coverage gap)
  • Severity (e.g., medium/low)
  • Location (which file, such as spec.md, plan.md, tasks.mmd)
  • Specific guidance for remediation

Examples of flagged issues (from the output)

Ambiguities

  • In spec.md / medium: unclear requirements (e.g., missing specific color values). Suggested remediation: define a palette or defer to a later design phase.

  • In plan.md / medium: performance/UX requirement like “fast initial load / smooth interactions” is not quantified. Suggested remediation: add measurable targets.

Inconsistencies

  • In tasks / low: reference confusion involving T02 vs T06 related to date utils.ts. Suggested remediation: clarify that T002 is installation vs T006 utilities.

  • In tasks / specification-related: business logic tasks T11–T13 lack specific file paths. Suggested remediation: add paths or guide the model to find goals.ts / split service files.

Coverage gaps

  • For a functional requirement like FR015, local storage error handling has no task to implement the error message/UI component. The report notes most requirements do have corresponding tasks, but this one is missing.

Metrics and next actions

Reported metrics (example from demo)

  • 19 functional requirements
  • 22 implementation tasks
  • ~95% coverage
  • 2 ambiguities

Recommendations included

  • Proceed with implementation despite minor issues (as suggested by the analysis output)
  • Optionally, the system asks whether to suggest concrete remediation edits for the top three issues, including the possibility to edit files/spec/plan/tasks before implementing in the next lesson.

Main speakers / sources

  • Main speaker: the presenter (author of the walkthrough) demonstrating GitHub Spec Kit usage.
  • Source being discussed: GitHub Copilot / coding agent behavior driven by the Spec Kit commands—specifically the /analyze prompt.

Original video