- Use full model names (claude-haiku-4-5, etc.) in create-capability
- Add allowed-tools to gitea skill for tea/jq commands
- Set default model to opus in settings
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- tea issues output has title on line 2, not line 1
- Update sed command to extract from correct line
- Fixes branches being named "issue-N-" or "issue-N-untitled"
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add step 5 to merge approved PRs using tea pulls merge --style rebase
- Clean up branch after merge with tea pulls clean
- Update role description and outputs to reflect merge responsibility
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Added explicit instructions to keep PR comments extremely brief:
- Maximum 3-4 bullet points
- One line per bullet
- Just state what was fixed
- No verbose explanations
- No code snippets
- No apologizing or thanking
Before: Long, verbose comments explaining every change in detail
After: "Fixed review feedback ✓
- Fixed error handling
- Added null checks
- Updated tests"
Updated:
- Added step 6: Post Concise Comment
- Added format examples (good vs bad)
- Added "Keep comments concise" to Guidelines
- Updated description and Your Role section
- Emphasized in Tips section
Co-Authored-By: Claude Code <noreply@anthropic.com>
Creates new user-invocable skill for fixing PRs based on review feedback:
- Takes multiple PR numbers as arguments
- Creates isolated fix worktrees for each PR
- Spawns pr-fixer agents in parallel
- Event-driven result handling
- Addresses review comments autonomously
- Commits and pushes fixes to Gitea using tea
- Shows detailed summary of all fixes
Uses tea and gitea skill (not gh). Pattern matches spawn-issues and
spawn-pr-reviews: Haiku orchestrator with allowed-tools.
Completes the spawn trilogy:
- spawn-issues: full workflow (implement → review → fix)
- spawn-pr-reviews: review only (read-only)
- spawn-pr-fixers: fix only (based on feedback)
Co-Authored-By: Claude Code <noreply@anthropic.com>
Creates new user-invocable skill for reviewing PRs:
- Takes multiple PR numbers as arguments
- Creates isolated review worktrees for each PR
- Spawns code-reviewer agents in parallel
- Event-driven result handling
- Posts review comments to Gitea using tea
- Read-only operation (no auto-fixes)
Uses tea and gitea skill (not gh), avoiding conflicts with base
system instructions.
Pattern matches spawn-issues: Haiku orchestrator with allowed-tools
for bash execution.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Removed outdated skills that haven't been migrated from old/ directory:
- /manifesto, /vision, /work-issue, /dashboard, /review-pr
- /create-issue, /retro, /plan-issues, /groom
Added current user-invocable skills:
- /vision-to-backlog
- /create-milestones
- /spawn-issues
- /create-capability
- /capability-writing
This prevents confusion when users try non-existent skills, which could
cause the system to fall back to base instructions (using gh instead of tea).
Co-Authored-By: Claude Code <noreply@anthropic.com>
The spawn-issues skill was missing the allowed-tools field that was present
in the old version. Without this field, the skill cannot execute bash commands,
causing permission errors when trying to create worktrees or call scripts.
Added: allowed-tools: Bash, Task, Read, TaskOutput
Co-Authored-By: Claude Code <noreply@anthropic.com>
Users were confused by ./scripts/ references, thinking they needed to copy
scripts into their project. Scripts are in ~/.claude/skills/worktrees/scripts/
and should be referenced with full paths.
Changes:
- Updated spawn-issues to use full script paths
- Updated worktrees skill with full paths in all examples
- Fixed gitea model name to claude-haiku-4-5
- Added tools list to issue-worker agent
Co-Authored-By: Claude Code <noreply@anthropic.com>
Restructure steps 7-8 to be clearer and more efficient:
- Merged "Assign Issues" and "Apply Labels" into single step
- Explicit nested loop structure: milestone → issues in that milestone
- Process one milestone at a time
- Combine milestone assignment + labels in single tea command
- Added clear examples
- Prevents confusion about looping and when to move on
Before: Separate loops for milestone assignment and label application
After: Single pass through milestones, process all issues per milestone
Co-Authored-By: Claude Code <noreply@anthropic.com>
Clarify that after user approves issues at decision gate, workflow should
automatically proceed to create all issues in Gitea without waiting for
another prompt.
Changes:
- Step 13: Clear yes/no question "Ready to create these issues in Gitea?"
- Step 14: Marked as "automatic after approval"
- Guidelines: Added "Automatic execution after approval" section with example
Prevents workflow from stopping and requiring user to explicitly request
issue creation after already approving the backlog.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Add capability for organizing backlog into shippable business capabilities
using value-based milestones (not time-based phases).
Components:
- milestone-planning skill: Value-based framework, vertical slice test, one active milestone
- create-milestones skill: Orchestrator (Haiku) for analyzing and grouping issues
- milestone-planner agent: Groups issues into capabilities autonomously (Haiku)
Core Principles:
- Milestone = shippable business capability (not phase)
- One active milestone at a time (preserves focus)
- 5-25 issues per milestone (right-sized)
- Value labels: value/high, value/medium, value/low
- Risk labels: risk/high (optional)
- Vertical slice test (can be demoed independently)
- No dates (capability-based, not time-based)
Workflow: /create-milestones reads existing Gitea issues → analyzes capability
boundaries → groups into milestones → creates in Gitea → assigns issues →
applies labels → user manually activates ONE milestone
Co-Authored-By: Claude Code <noreply@anthropic.com>
Split gitea skill into main file and reference documentation.
Main SKILL.md now focuses on core commands (154 lines, down from 201),
with setup/auth and CI/Actions moved to reference files.
Co-Authored-By: Claude Code <noreply@anthropic.com>
Remove top-level model override to allow per-skill/agent model configuration.
Reorder sections for consistency.
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
Claude Code has unified commands into skills with the user-invocable
frontmatter field. This migration:
- Converts 20 commands to skills with user-invocable: true
- Consolidates docs into single writing-capabilities.md
- Rewrites capability-writing skill for unified model
- Updates CLAUDE.md, Makefile, and other references
- Removes commands/ directory
Skills now have two types:
- user-invocable: true - workflows users trigger with /name
- user-invocable: false - background knowledge auto-loaded
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Resolves issue #86 by having the spawn-issues orchestrator create worktrees
upfront and pass the worktree paths to agents, instead of having agents
create their own worktrees in sibling directories outside the sandbox.
Changes:
- spawn-issues orchestrator creates all worktrees before spawning agents
- issue-worker, pr-fixer, code-reviewer accept optional WORKTREE_PATH
- When WORKTREE_PATH is provided, agents work directly in that directory
- Backward compatible: agents still support creating their own worktrees
if WORKTREE_PATH is not provided
- Orchestrator handles all worktree cleanup after agents complete
- Eliminates permission denied errors from agents trying to access
sibling worktree directories
This ensures agents operate within their sandbox while still being able to
work with isolated git worktrees for parallel implementation.
Closes#86
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Set explicit model preferences to optimize for speed vs capability:
- haiku: 11 commands, 2 agents (issue-worker, pr-fixer), 10 skills
Fast execution for straightforward tasks
- sonnet: 4 commands (groom, improve, plan-issues, review-pr),
1 agent (code-reviewer)
Better judgment for analysis and review tasks
- opus: 2 commands (arch-refine-issue, arch-review-repo),
1 agent (software-architect)
Deep reasoning for architectural analysis
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add linter detection logic that checks for common linter config files
(ESLint, Ruff, Flake8, Pylint, golangci-lint, Clippy, RuboCop)
- Add instructions to run linter on changed files only
- Add "Lint Issues" section to review output format
- Clearly distinguish lint issues from logic/security issues
- Document that lint issues alone should not block PRs
Closes#25
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Introduces a new command that guides users through creating capabilities
for the architecture repository. The command analyzes user descriptions,
recommends appropriate component combinations (skill, command, agent),
gathers necessary information, generates files from templates, and presents
them for approval before creation.
Closes#75
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add model: sonnet to issue-worker agent (balanced for implementation)
- Add model: sonnet to pr-fixer agent (balanced for feedback iteration)
- Add model: haiku to /dashboard command (read-only display)
- Add model: haiku to /roadmap command (read-only categorization)
- Document rationale for each model selection in frontmatter comments
Closes#72
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Creates a lighter-weight PR creation flow for when you're already on a
branch with commits. Features:
- Auto-generates title from branch name or commits
- Auto-generates description summarizing changes
- Links to related issue if branch name contains issue number
- Triggers code-reviewer agent after PR creation
Closes#19
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add new section to display recent workflow runs from tea actions runs
- Show status indicators: [SUCCESS], [FAILURE], [RUNNING], [PENDING]
- Highlight failed runs with bold formatting for visibility
- Gracefully handle repos without CI configured
- Include example output format for clarity
Closes#20
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The planning process previously jumped directly from understanding a feature
to breaking it down into issues. This led to proposing issues without first
understanding the user's actual workflow and where the gaps are.
Added a discovery phase that requires walking through:
- Who is the specific user
- What is their goal
- Step-by-step workflow to reach the goal
- What exists today
- Where the workflow breaks or has gaps
- What's the MVP
Issues are now derived from workflow gaps rather than guessing.
Closes#29
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add streamlined commit workflow that analyzes staged changes and
generates conventional commit messages (feat:, fix:, etc.) with
user approval before committing.
Closes#18
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Creates a skill that teaches how to design and create capabilities
(skill + command + agent combinations) for the architecture repository.
Includes:
- Component templates for skills, commands, and agents
- Decision tree and matrix for when to use each component
- Model selection guidance (haiku/sonnet/opus)
- Naming conventions and anti-patterns to avoid
- References to detailed documentation in docs/
- Checklists for creating each component type
Closes#74
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Architecture Beliefs section to manifesto with outcome-focused beliefs:
auditability, business language in code, independent evolution, explicit over implicit
- Create software-architecture.md as human-readable documentation
- Enhance software-architecture skill with beliefs→patterns mapping (DDD, Event
Sourcing, event-driven communication) and auto-trigger description
- Update work-issue command to reference skill and check project architecture
- Update issue-worker agent with software-architecture skill
- Add Architecture section template to vision-management skill
The skill is now auto-triggered when implementing, reviewing, or planning
architectural work. Project-level architecture choices go in vision.md.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The code reviewer prompt was minimal and didn't specify worktree setup,
causing parallel reviewers to interfere with each other by checking out
different branches in the same directory.
Changes:
- Add worktree setup/cleanup to code reviewer prompt (like issue-worker/fixer)
- Add branch tracking to issue state
- Add note about passing branch name to reviewers
- Expand reviewer prompt with full review process
This ensures each reviewer works in isolation at:
../<repo>-review-<pr-number>
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Creates a new command that refines issues with architectural perspective
by spawning the software-architect agent to analyze the codebase before
proposing implementation guidance. The command:
- Fetches issue details and spawns software-architect agent
- Analyzes existing patterns and affected components
- Identifies architectural concerns and dependencies
- Proposes refined description with technical notes
- Allows user to apply, edit, or skip the refinement
Closes#59
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add software architecture review as a standard part of PR review process:
- Reference software-architecture skill for patterns and checklists
- Spawn software-architect agent for architectural analysis
- Add checks for pattern consistency, dependency direction, breaking changes,
module boundaries, and error handling
- Structure review output with separate Code Review and Architecture Review
sections
Closes#60
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Creates a new command that spawns the software-architect agent to perform
comprehensive architecture audits. The command analyzes directory structure,
package organization, patterns, anti-patterns, dependencies, and test coverage,
then presents prioritized recommendations with a health score.
Closes#58
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace polling loop with task-notification based orchestration.
Background tasks send notifications when complete - no need to poll.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Create the software-architect agent that performs deep architectural
analysis on codebases. The agent:
- References software-architecture skill for patterns and checklists
- Supports three analysis types: repo-audit, issue-refine, pr-review
- Analyzes codebase structure and patterns
- Applies architectural review checklists from the skill
- Identifies anti-patterns (god packages, circular deps, etc.)
- Generates prioritized recommendations (P0-P3)
- Returns structured ARCHITECT_ANALYSIS_RESULT for calling commands
Closes#57
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Creates the foundational skill that encodes software architecture
best practices, review checklists, and patterns for Go and generic
architecture guidance.
Closes#56
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Each issue flows independently through: implement → review → fix → review
- Don't wait for all workers before starting reviews
- Print status update as each step completes
- Poll loop checks all tasks, advances each issue independently
- State machine: implementing → reviewing → fixing → approved/failed
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
