* fix(execute-phase): pass file paths to subagents instead of content
The --include flag added in fa81821 caused orchestrator context bloat
by reading STATE, config, and plan files into the orchestrator's context,
then embedding all content in Task() prompts.
With multiple plans, this consumed 50-60%+ of context before execution.
Fix: Pass file paths only. Subagents read files themselves in their
fresh 200k context. Orchestrator stays lean (~10-15% as intended).
Fixes#479
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix: respect commit_docs=false in execute-plan and debugger workflows
Two code paths bypassed the commit_docs configuration check, causing
.planning files to be intermittently committed when commit_docs=false:
1. execute-plan.md: update_codebase_map step ran `git add .planning/codebase/*.md`
unconditionally — now gated behind commit_docs check
2. gsd-debugger.md: used `git add -A` which stages .planning/ files — replaced
with explicit individual file staging and proper commit_docs conditional
Fixes#478https://claude.ai/code/session_013yS1F2VR3Jn2pdwqr5NuDo
* fix: route all .planning commits through gsd-tools.js CLI
Instead of wrapping direct git commands in markdown conditionals,
both bypass paths now use gsd-tools.js commit which has the
commit_docs check built in:
1. execute-plan.md: uses `gsd-tools.js commit --amend` for codebase
map updates (new --amend flag added to CLI)
2. gsd-debugger.md: code commit uses direct git (no .planning files),
planning docs commit uses gsd-tools.js commit
Also added --amend support to gsd-tools.js commit command so the
execute-plan codebase map step can amend the previous metadata commit.
Fixes#478https://claude.ai/code/session_013yS1F2VR3Jn2pdwqr5NuDo
* docs: update reference docs to use gsd-tools.js CLI for all .planning commits
Reference documentation showed direct git add/commit patterns for
.planning files, which agents copy-paste and bypass the commit_docs
check. Updated all three reference files to show gsd-tools.js commit
as the canonical pattern:
- git-planning-commit.md: replaced manual bash conditionals with CLI
- git-integration.md: replaced direct git add/commit in initialization,
plan-completion, and handoff examples
- planning-config.md: replaced conditional git example with CLI call
https://claude.ai/code/session_013yS1F2VR3Jn2pdwqr5NuDo
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
- Add "Squash merge" as recommended option at complete-milestone
- Keep "Merge with history" (--no-ff) as alternative
- Document merge options in planning-config reference
Improves on #298 based on community feedback from @oojacoboo.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add configurable git branching with three strategies:
- none (default): commit to current branch
- phase: create branch per phase (gsd/phase-{N}-{slug})
- milestone: create branch per milestone (gsd/{version}-{slug})
Changes:
- planning-config.md: add git.branching_strategy and templates
- execute-phase.md: handle branch creation based on strategy
- settings.md: add branching strategy to settings UI
- complete-milestone.md: handle branch merging at milestone end
* docs: enforce automation-first checkpoint verification
Checkpoints should never ask users to run CLI commands that Claude Code
can execute. This update reinforces the automation-first principle:
Key changes:
- Add golden rules: Claude runs CLI, users only visit URLs
- Add dev server automation patterns (start before checkpoint)
- Add environment variable CLI patterns (Convex, Vercel, etc.)
- Add anti-patterns: asking user to run npm, add dashboard env vars
- Update all examples to show Claude starting servers
- Add comprehensive "Never Ask Users To" and "Users Only Do" lists
- Update gsd-executor with pre-checkpoint automation requirements
The core principle: if Claude CAN automate it, Claude MUST automate it.
Users only do what requires human judgment (visual verification, UX).
* refactor: DRY checkpoint automation with server lifecycle and error handling
Changes:
- checkpoints.md is now single source of truth for automation-first patterns
- Added server lifecycle protocol (start, port conflicts, cleanup)
- Added CLI installation handling (auto-install matrix)
- Added pre-checkpoint failure handling (fix before checkpoint)
- Removed ~93 lines of duplication from verification-patterns.md
- Replaced inline examples in phase-prompt.md with references
- Slimmed gsd-executor.md checkpoint section to reference checkpoints.md
Net effect: -23 lines while adding 3 new capabilities (server lifecycle,
CLI install, error handling). Single place to update automation patterns.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Adds switchable model profiles (quality/balanced/budget) to control
which Claude model each agent uses, balancing quality vs token spend.
New files:
- get-shit-done/references/model-profiles.md - profile lookup table
- commands/gsd/set-profile.md - runtime profile switching
Updated orchestrators to resolve model before spawning agents:
- new-project, new-milestone (7 calls each)
- plan-phase (4 calls)
- execute-phase (3 calls)
- audit-milestone, debug, research-phase, quick (1-2 calls each)
- workflows: execute-phase, execute-plan, verify-work, map-codebase
Profile stored in .planning/config.json as model_profile.
Default: "balanced" (Opus for planning, Sonnet for execution).
Closes#160
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: add uncommitted planning mode
Add config option to keep planning docs local-only (not committed to git).
Useful for OSS contributions, client projects, or keeping planning private.
Config options in .planning/config.json:
- planning.commit_docs: true/false (default: true)
- planning.search_gitignored: true/false (default: false)
When commit_docs=false:
- All git operations for .planning/ files are skipped
- User should add .planning/ to .gitignore
- Planning system works normally, just not tracked
Updated files:
- config.json template: added planning section
- execute-plan.md: conditional git commits
- execute-phase.md: conditional git commits
- create-roadmap.md: conditional git commits
- help.md: documented new config options
- planning-config.md: reference doc for config behavior
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* feat: extend commit_docs check to all agents/commands/workflows
Add COMMIT_PLANNING_DOCS config check to all files that commit
.planning/ artifacts, ensuring consistent behavior when
commit_docs=false is set in config.json.
Updated:
- 5 agents (planner, executor, debugger, phase-researcher, synthesizer)
- 8 commands (add-todo, check-todos, execute-phase, new-milestone,
pause-work, plan-milestone-gaps, remove-phase, research-project)
- 7 workflows (complete-milestone, create-milestone, define-requirements,
diagnose-issues, discuss-phase, map-codebase, verify-work)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(progress): use Bash instead of Glob for .planning/ check
Glob respects .gitignore, so projects with gitignored .planning/
directories would fail with "No planning structure found."
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* fix(new-milestone): integrate full research/requirements/roadmap flow
Previously routed to non-existent /gsd:research-project and
/gsd:define-requirements commands. Now handles the full flow inline
like new-project does:
- Phase 7: Research Decision (spawns 4 milestone-aware researchers)
- Phase 8: Define Requirements (scopes features, creates REQUIREMENTS.md)
- Phase 9: Create Roadmap (continues phase numbering from previous milestone)
- Phase 10: Done
Key adaptations for milestones:
- Research focuses on NEW features only
- Requirements add to existing, don't start fresh
- Phase numbering continues from previous milestone
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
/gsd:execute-phase handles single and multiple plans.
All references updated to use execute-phase.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Commands load agent expertise directly via Task tool spawning.
Thin orchestrator pattern — agents have methodology baked in.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add ui-brand.md reference with visual patterns:
- Stage banners with GSD ► prefix
- Checkpoint boxes (62-char width)
- Status symbols vocabulary
- Next Up routing blocks
Update orchestrators to use branded output.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- principles.md: redirects to <philosophy> section
- plan-format.md: redirects to <plan_format> section
- scope-estimation.md: redirects to <scope_estimation> section
- goal-backward.md: redirects to <goal_backward> section
All content preserved in agents/gsd-planner.md
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The ~/.claude/skills/expertise/ system was personal tooling
that doesn't exist for other users. Removed from:
- create-roadmap workflow (detect_domain step)
- plan-phase workflow (domain loading)
- roadmap template (Domain Expertise section)
- phase-prompt template (domain frontmatter)
- plan-format reference (domain field)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Use double-line box for checkpoint headers with clear visual hierarchy:
- Progress and task info at top
- Content details in middle
- Separator line before unmissable "→ YOUR ACTION:" prompt
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Reduce principles.md from 158 to 73 lines:
- Remove duplicates (atomic_commits, tdd, deviation_rules)
- Remove version drag from claude_automates
- Keep core orientation: solo dev model, plans are prompts,
scope control, ship fast, anti-enterprise
Add principles.md to 9 core commands so Claude always
understands what GSD is.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Focus on what to extract, not what to avoid:
- What they're building
- Why it needs to exist
- Who it's for
- What done looks like
Remove version drag and redundant question categories.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Follow threads naturally instead of rigid step order
- Cleaner question techniques in questioning.md
- Background context checklist instead of forced checklist mode
- Updated next steps to show research flow
- execute-plan.md: Replace broken Task(resume=...) with fresh continuation agent
- execute-phase.md: Expand checkpoint handling with full continuation flow
- checkpoints.md: Restore ~500 lines of detailed examples and anti-patterns
Checkpoint presentation formats now documented for all three types.
Templates referenced: checkpoint-return.md, continuation-prompt.md
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Wave numbers now computed during plan-phase and stored in PLAN.md
frontmatter. Execute-phase reads wave directly instead of deriving
from depends_on at runtime.
- Add assign_waves step to plan-phase workflow
- Add wave field to frontmatter (plan-format, phase-prompt template)
- Simplify execute-phase: remove analyze_dependencies and group_into_waves
- Replace with group_by_wave that just reads frontmatter integers
Co-Authored-By: Claude <noreply@anthropic.com>
TDD requires 2-3 execution cycles (RED → GREEN → REFACTOR), each with
file reads, test runs, and potential debugging. This consumes 40-50%
of context for a single feature. Dedicated TDD plans ensure full
quality throughout the cycle.
- TDD plans use `type: tdd` frontmatter with `<feature>` element
- One feature per TDD plan (batching defeats the purpose)
- Target ~40% context usage for TDD plans
- Standard plans handle non-TDD work with 2-3 tasks
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Each task commits immediately after completion (feat/fix/test/refactor)
- Plan completion commits metadata only (docs: SUMMARY + STATE + ROADMAP)
- Git history becomes primary context source for future Claude sessions
- Better bisect granularity, failure recovery, and revert capabilities
Changes:
- execute-phase.md: Added <task_commit> protocol and per-task commit flow
- execute-plan.md: Updated commit rules and success criteria
- git-integration.md: New task-completion format, strategy rationale
- tdd.md: Aligned TDD commits with standard task commit pattern
- summary.md: Added Task Commits section with commit hashes
- README.md: Rewrote git history section with atomic commit benefits
- scope-estimation.md: Clarified commit counts per plan
- principles.md: Added <atomic_commits> section explaining strategy
Each plan now produces 3-4 commits (2-3 task commits + 1 metadata commit)
instead of single monolithic commit. Optimizes observability for solo
developer + Claude Code workflow.
Depth sets a ceiling on compression, not a floor on expansion.
Small phases stay small even in comprehensive mode.
- Comprehensive auth = 8 plans (work warrants it)
- Comprehensive "add favicon" = 1 plan (that's all it is)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Adds depth setting (quick/standard/comprehensive) to control how many
phases and plans get created. Tasks per plan stays constant at 2-3.
- Quick: 3-5 phases, 1-3 plans each
- Standard: 5-8 phases, 3-5 plans each
- Comprehensive: 8-12 phases, 5-10 plans each
Depth increases plan COUNT, never plan SIZE. More plans = more
thoroughness with same quality per plan.
Files changed:
- new-project.md: depth question after mode
- config.json: depth field added
- create-roadmap.md: depth-aware phase guidance
- plan-phase.md: depth-aware plan splitting
- scope-estimation.md: depth calibration section
- create-milestone.md: removed hardcoded 3-6 limit
- new-milestone.md: removed hardcoded 3-6 limit
- roadmap.md template: depth-aware guidance
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add references/tdd.md with detection heuristics, red-green-refactor cycle
- Add tdd_annotation section to plan-format.md with task structure
- Expand plan-phase.md with TDD detection and annotation guidance
- Add tdd_execution section to execute-phase.md with commit patterns
- TDD-aware task routing in execute step
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Consolidated checkpoints.md with automation reference (287 lines from 1,121)
- Compressed plan-format.md to reference card (127 lines from 397)
- Deleted cli-automation.md (merged into checkpoints.md)
- Updated plan-phase.md to remove stale reference
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add continuation-format.md reference defining the standard format
- Replace "## To Continue" with "## ▶ Next Up"
- Show plan/phase name + description instead of just command paths
- Add `/clear` explanation: "fresh context window"
- Change "Other options" to "Also available"
- Add /gsd:research-phase to phase planning options
- Add visual separators and completion headers
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove fixed 3-6 phase constraint and time-based sizing (1-3 days).
Phase count now emerges from deliverable coherence—simple projects
get fewer phases, complex projects get more.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
All questioning workflows now require AskUserQuestion tool for every
question, not just decision gates. Inline text questions are prohibited.
Updated: questioning.md, new-project.md, discuss-phase.md,
discuss-milestone.md, and their workflow files.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Decision gate now explicitly uses AskUserQuestion tool, enabling
the loop behavior where users can keep exploring before finalizing.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Research is now optimized for Claude Code implementation quality:
- PROJECT.md-driven (features, constraints, questions, decisions)
- High-confidence only (no low-quality padding)
- 3 focused files: stack.md, implementation.md, risks.md
- Code examples with current vs deprecated patterns
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Replace interrogation-style domain checklist with conversation arc
- Focus on following user's thread, not walking a checklist
- Add good vs bad examples to guide natural questioning
- Update new-project to use new philosophy
Questioning now helps users discover and articulate their vision
through collaborative thinking instead of form-filling.
- Created comprehensive prompts for ecosystem, architecture, pitfalls, standards
- Updated workflow with required_reading and concrete Task tool examples
- Batch ordering: ecosystem+architecture first, pitfalls+standards second
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add /gsd:research-phase command for comprehensive ecosystem research
- Rename FINDINGS.md → DISCOVERY.md for clarity
- Rename research-phase.md → discovery-phase.md (shallow "which library")
- New research-phase.md for deep "how experts build this" research
- Creates RESEARCH.md with stack, patterns, pitfalls, don't-hand-roll
- plan-phase loads RESEARCH.md when present
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- New <test_driven_when_beneficial> principle in references/principles.md
- TDD assessment step in plan-phase.md break_into_tasks
- Pragmatic approach: use TDD when behavior is definable upfront
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
