Research now loads concrete requirements (e.g., "AUTH-02: User receives
email verification") to inform research domains, not just high-level
roadmap descriptions.
Co-Authored-By: Claude <noreply@anthropic.com>
- Remove phantom status.md command (background agent model abandoned)
- Remove agent-history.md template (unused)
- Remove _archive/ directory
- Add narration to execute-phase (describe what's being built before/after waves)
- Update new-project to offer define-requirements as fast path
- Make define-requirements work without research (gather through questioning)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Subagents can't determine if phase is complete. The orchestrator
(execute-plan and execute-phase commands) runs in main context,
can count plans vs summaries, and updates REQUIREMENTS.md when
phase completes.
- execute-plan: Step 2.5 updates requirements when summaries = plans
- execute-phase: Step 6 updates requirements (phase always complete)
- Removed duplicate from workflow (subagent can't run it anyway)
When all plans in a phase complete:
1. Look up phase's requirements from ROADMAP.md
2. Update REQUIREMENTS.md traceability table (Pending → Complete)
3. Include in metadata commit
Closes the loop on requirement tracking.
- Roadmap template now includes Requirements: field for each phase
- plan-phase loads REQUIREMENTS.md and extracts phase-specific requirements
- define-requirements shows full list inline before commit (not just counts)
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
Adds /gsd:research-project to research domain ecosystem before creating roadmap.
Spawns parallel agents to investigate stack, features, architecture, and pitfalls.
New files:
- commands/gsd/research-project.md
- get-shit-done/workflows/research-project.md
- get-shit-done/templates/research-project/ (5 templates)
Modified:
- new-project.md: offers research vs direct roadmap options
- create-roadmap.md: loads research if exists
- workflows/create-roadmap.md: uses research to inform phases
Co-Authored-By: Claude <noreply@anthropic.com>
- 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>
Remove the global ISSUES.md deferred enhancement tracking system.
- Delete /gsd:consider-issues command (never used)
- Delete issues.md template (never instantiated)
- Remove Rule 5 from deviation rules (never triggered)
- Remove all ISSUES.md, ISS-XXX, and "deferred issues" references
- Update STATE.md to track pending todos instead
The ISSUES.md system was designed to capture non-critical enhancements
during plan execution via "Rule 5", but it never fired in practice
across 100+ projects. The system added ~350 lines of dead code.
The /gsd:add-todo and /gsd:check-todos system serves the same purpose
and is actually used.
Note: UAT *-ISSUES.md files (per-plan, created by /gsd:verify-work)
are unaffected - those are a separate, active system.
Closes#56
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Debug command now orchestrates: gather symptoms in main context, spawn
investigation subagent with fresh 200k context
- Subagent template unified for both /gsd:debug (find_and_fix) and
diagnose-issues (find_root_cause_only) flows via goal flag
- Checkpoint behavior enables subagent to pause for user input
(human-verify, human-action, decision) with continuation agents
- Structured return formats: ROOT CAUSE FOUND, DEBUG COMPLETE,
INVESTIGATION INCONCLUSIVE, CHECKPOINT REACHED
- diagnose-issues updated to match new template placeholders and returns
Co-Authored-By: Claude <noreply@anthropic.com>
Define DEBUG_DIR=.planning/debug at top of debug-related files.
Reference ${DEBUG_DIR} throughout to reduce path typo risk.
Co-Authored-By: Claude <noreply@anthropic.com>
- Debug files from UAT use same naming as regular debug (slug only)
- UAT.md tracks link via debug_session field
- plan-fix actually invokes execute-plan when user selects it
Co-Authored-By: Claude <noreply@anthropic.com>
Diagnosis always produces better fixes and runs in parallel anyway.
Removing the prompt reduces cognitive load.
Co-Authored-By: Claude <noreply@anthropic.com>
After UAT finds issues, spawn parallel debug agents to investigate
root causes before planning fixes. Each agent investigates one issue
with symptoms pre-filled from UAT, finds the root cause, and returns
diagnosis.
New files:
- workflows/diagnose-issues.md: Orchestrator for parallel debug agents
- templates/debug-subagent-prompt.md: Prompt template for debug subagents
Modified:
- workflows/debug.md: Add symptoms_prefilled and diagnose-only modes
- workflows/verify-work.md: Offer diagnosis step after issues found
- templates/UAT.md: Add root_cause and debug_session fields
- commands/gsd/plan-fix.md: Use root causes for targeted fix planning
Flow: UAT → diagnose (parallel) → plan-fix (with root causes) → execute
Co-Authored-By: Claude <noreply@anthropic.com>
- One test at a time instead of full checklist upfront
- Plain text responses instead of AskUserQuestion forms
- Severity inferred from description, never asked
- Persistent UAT.md survives /clear (like debug workflow)
- Single file per phase replaces per-plan ISSUES.md
- Updated plan-fix to read from UAT.md
Co-Authored-By: Claude <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>
- Fix brownfield example to use execute-phase instead of execute-plan
- Remove /gsd:resume-task command (relies on broken Task resume)
- Checkpoint continuation now uses fresh agents, not resume
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Resume fails when subagents use parallel tool calls due to Claude Code
serialization bug (consecutive assistant entries with same message ID).
Solution: Subagents return structured checkpoint state, orchestrator
spawns fresh agent with continuation-prompt template instead of resuming.
New files:
- templates/checkpoint-return.md: Structured format with completed tasks table
- templates/continuation-prompt.md: Template for spawning continuation agent
Updated:
- templates/subagent-task-prompt.md: Reference checkpoint-return, remove resume language
- workflows/execute-phase.md: Replace Task(resume=id) with fresh agent spawn
- workflows/execute-plan.md: Update checkpoint_return_for_orchestrator step
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
execute-plan.md command now spawns subagent instead of executing directly:
- Loads only subagent-task-prompt template (~100 lines vs ~2200)
- Subagent loads full execute-plan workflow, summary, checkpoints, tdd
- Handles checkpoint returns with resume flow
- ~80% context reduction for orchestrator
Also updated subagent-task-prompt.md description to clarify it's used by
both execute-phase (parallel) and execute-plan (single) orchestrators.
Co-Authored-By: Claude <noreply@anthropic.com>
Scope boundaries are implicit from the roadmap. Asking about them
creates the sensation of scope creep and interrogates the user about
constraints they didn't mention.
Co-Authored-By: Claude <noreply@anthropic.com>
Restored context-budget reasoning for why TDD features get dedicated plans:
- TDD requires 2-3 execution cycles consuming 50-60% context
- Test framework setup handled in first TDD plan's RED phase
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- execute-plan.md: Add checkpoint_return_for_orchestrator step
explaining how to return at checkpoints when spawned via Task tool
- subagent-task-prompt.md: Add checkpoint_behavior and completion_format
sections to guide agents on returning for checkpoints
Tested: Task resume works - agent pauses at checkpoint, returns with
details, orchestrator presents to user, resumes with Task(resume=id).
Parallel agents each get unique agent_id for independent resume.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Shows table of all active debug sessions with status/hypothesis/next action
- Reply with number to resume, or describe new issue
- No AskUserQuestion for session selection - inline flow
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove confusing `parallelizable` boolean field. It was redundant -
execute-phase already computes parallelizability from `depends_on`
and file conflicts.
Changes:
- Remove `parallelizable: true|false` from plan frontmatter
- Rename `files_exclusive` to `files_modified` (clearer intent)
- Simplify execute-phase dependency detection logic
- Update phase-prompt template and plan-phase workflow
Parallelization is now fully automatic based on:
- `depends_on: []` (empty = independent)
- `files_modified: [...]` (no overlap = can parallelize)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Routes to batch execution when 2+ plans exist, with execute-plan
as alternative for interactive mode.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Context isolation is the primary goal, not parallelization.
Even sequential plans spawn separate subagents for fresh 200k context.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add parallel-aware plan example with vertical slice pattern
- Add sequential plan example with explicit dependencies
- Document key differences between parallel and sequential plans
- Update categorization table to show frontmatter-aware criteria
- Add categorization priority documentation
- Update calculate_wave() to check frontmatter depends_on first
- Handle parallelizable: false as implicit Wave 2+ assignment
- Fall back to inferred requires only when no frontmatter
- Add [frontmatter] annotation in wave output for plans using new fields
- Add PLAN_PARALLELIZABLE, PLAN_DEPENDS_ON, PLAN_FILES_EXCLUSIVE arrays
- Check for parallelizable, depends_on, files_exclusive in plan frontmatter
- Use frontmatter values directly when present
- Fall back to inference logic for old plans without frontmatter
- Document dependency detection priority
- Include parallelizable, depends_on, files_exclusive in frontmatter
- Rules for parallelizable field determination
- Parallel-aware context section guidance
- Avoid reflexive SUMMARY references when parallel
- File ownership analysis per task group
- Detect unnecessary inter-plan dependencies
- Restructure for vertical slices when beneficial
- Mark plans with parallelization frontmatter values
- Reads .planning/config.json parallelization settings early
- Documents planning behavior when parallel enabled vs disabled
- Placed after load_project_state, before load_codebase_context
- Default: assume parallelization enabled for new projects
- parallelizable: marks plan as safe for concurrent execution
- depends_on: explicit plan dependencies
- files_exclusive: conflict detection for parallel agents
- Added frontmatter_guidance section with usage rules
- Updated good_examples with parallel/sequential examples
- Sequential execution example with all fields
- Plan-level parallel execution example
- Queued with dependency example
- Document parallel_group format conventions
- Command now loads both workflows (execution + orchestration)
- Agents commit tasks normally (removed DON'T COMMIT instruction)
- Orchestrator only commits metadata after all agents complete
- No duplication - single source of truth for execution logic
