Skip to content

Add feature-spec skill for pre-implementation documentation - #8

Open
zhumo wants to merge 49 commits into
masterfrom
feature-spec-workflow
Open

Add feature-spec skill for pre-implementation documentation#8
zhumo wants to merge 49 commits into
masterfrom
feature-spec-workflow

Conversation

@zhumo

@zhumo zhumo commented Jan 8, 2026

Copy link
Copy Markdown
Owner

Summary

  • Adds a new /feature-spec skill that generates comprehensive feature documentation
  • Creates PRD through PM-style interview workflow
  • Auto-generates Technical Architecture and Testing documents from the PRD
  • Documents saved to projects/{branch-name}/ folder

Files Added

  • claude/skills/feature-spec/SKILL.md - Main skill definition and workflow
  • claude/skills/feature-spec/templates/prd.md - PRD template
  • claude/skills/feature-spec/templates/architecture.md - Architecture template
  • claude/skills/feature-spec/templates/testing.md - Testing template
  • claude/skills/feature-spec/references/interview-guide.md - Interview question bank

Test plan

  • Run symlink script to deploy skill to ~/.claude/
  • Invoke /feature-spec "test feature" and verify interview starts
  • Complete interview and verify PRD generates correctly
  • Verify Architecture doc generates from PRD
  • Verify Testing doc generates from PRD + Architecture
  • Check all docs saved to projects/{branch}/ folder

🤖 Generated with Claude Code

zhumo and others added 30 commits January 7, 2026 18:49
Creates a PM-style interview workflow that generates three documents:
- PRD (Product Requirements Document) via interview
- Technical Architecture doc from PRD
- Testing doc with unit and integration test plans

Documents are saved to projects/{branch-name}/ in the repo root.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Rename PRD to Requirements document
- Remove product-focused questions (personas, success metrics)
- Add edge cases, error handling, concurrency questions
- Add data persistence, scale, validation questions
- Add integration points, backwards compatibility questions
- Focus on concrete behavior and technical constraints

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Use exit code 1 with stderr output so users can see usage reporting
status. Exit code 0 hides all output, while exit code 2 would inject
messages into the model context unnecessarily.

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Remove /add-dir and cd instructions that don't work mid-session.
Instead, output a copy-pasteable command for the user to run in a new terminal.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* get id of latest message in server then send up all messages after that

* simplify the code

* only spit out simplified error

* catch the other error

* send to stderr

* format prettier error

* fix logic
SubagentStop provides agent_transcript_path for the subagent's transcript,
not transcript_path which is the main session. Use agent_id as session_id
for subagent usage tracking.

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant