gsd-codex-cli 1.20.0

package/.codex/prompts/gsd-add-phase.md

@@ -0,0 +1,44 @@
+---
+description: Add phase to end of current milestone in roadmap
+argument-hint: <description>
+---
+
+
+
+## Objective
+Add phase to end of current milestone in roadmap
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse <description> from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init phase-op "0" --raw
+
+3. Load .claude/get-shit-done/workflows/add-phase.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+
+

package/.codex/prompts/gsd-add-todo.md

@@ -0,0 +1,43 @@
+---
+description: Capture idea or task as todo from current conversation context
+argument-hint: [optional description]
+---
+
+
+
+## Objective
+Capture idea or task as todo from current conversation context
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse [optional description] from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init todos $ARG --raw
+
+3. Load .claude/get-shit-done/workflows/add-todo.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-audit-milestone.md

@@ -0,0 +1,43 @@
+---
+description: Audit milestone completion against original intent before archiving
+argument-hint: "[version]"
+---
+
+
+
+## Objective
+Audit milestone completion against original intent before archiving
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse "[version]" from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init milestone-op --raw
+
+3. Load .claude/get-shit-done/workflows/audit-milestone.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-check-todos.md

@@ -0,0 +1,43 @@
+---
+description: List pending todos and select one to work on
+argument-hint: [area filter]
+---
+
+
+
+## Objective
+List pending todos and select one to work on
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse [area filter] from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init todos $ARG --raw
+
+3. Load .claude/get-shit-done/workflows/check-todos.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-complete-milestone.md

@@ -0,0 +1,43 @@
+---
+description: Archive completed milestone and prepare for next version
+argument-hint: <version>
+---
+
+
+
+## Objective
+Archive completed milestone and prepare for next version
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse <version> from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init execute-phase "1" --raw
+
+3. Load .claude/get-shit-done/workflows/complete-milestone.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-debug.md

@@ -0,0 +1,46 @@
+---
+description: Systematic debugging with persistent state across context resets
+argument-hint: [issue description]
+---
+
+
+
+## Objective
+Systematic debugging with persistent state across context resets
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse [issue description] from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js state load --raw
+
+3. Execute the debug flow defined in this command (no separate workflow file exists in the gsd commands set).
+4. Translate each Task(...) in the debug flow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. Checkpoint handling:
+   - If an active debug session exists and no new issue was provided, list open sessions and ask the user which to resume.
+   - If a checkpoint is returned by the debugger, summarize the checkpoint and continue only after user confirmation.
+8. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-discuss-phase.md

@@ -0,0 +1,43 @@
+---
+description: Gather phase context through adaptive questioning before planning
+argument-hint: "<phase>"
+---
+
+
+
+## Objective
+Gather phase context through adaptive questioning before planning
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse "<phase>" from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init phase-op $ARG --raw
+
+3. Load .claude/get-shit-done/workflows/discuss-phase.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-execute-phase.md

@@ -0,0 +1,43 @@
+---
+description: Execute all plans in a phase with wave-based parallelization
+argument-hint: "<phase-number> [--gaps-only]"
+---
+
+
+
+## Objective
+Execute all plans in a phase with wave-based parallelization
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse "<phase-number> [--gaps-only]" from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init execute-phase $PHASE --raw
+
+3. Load .claude/get-shit-done/workflows/execute-phase.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-help.md

@@ -0,0 +1,43 @@
+---
+description: Show available GSD commands and usage guide
+argument-hint: [none]
+---
+
+
+
+## Objective
+Show available GSD commands and usage guide
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse [none] from the user input.
+2. Run init:
+# No explicit init.
+
+3. Load .claude/get-shit-done/workflows/help.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-insert-phase.md

@@ -0,0 +1,43 @@
+---
+description: Insert urgent work as decimal phase (for example 72.1) between existing phases
+argument-hint: <after> <description>
+---
+
+
+
+## Objective
+Insert urgent work as decimal phase (for example 72.1) between existing phases
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse <after> <description> from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init phase-op $AFTER --raw
+
+3. Load .claude/get-shit-done/workflows/insert-phase.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-list-phase-assumptions.md

@@ -0,0 +1,43 @@
+---
+description: Surface Claude's assumptions about a phase approach before planning
+argument-hint: "[phase]"
+---
+
+
+
+## Objective
+Surface Claude's assumptions about a phase approach before planning
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse "[phase]" from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init phase-op $ARG --raw
+
+3. Load .claude/get-shit-done/workflows/list-phase-assumptions.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-map-codebase.md

@@ -0,0 +1,43 @@
+---
+description: Analyze codebase with parallel mapper agents to produce .planning/codebase/ documents
+argument-hint: "[optional: specific area to map, for example api or auth]"
+---
+
+
+
+## Objective
+Analyze codebase with parallel mapper agents to produce .planning/codebase/ documents
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse "[optional: specific area to map, for example api or auth]" from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init map-codebase --raw
+
+3. Load .claude/get-shit-done/workflows/map-codebase.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-new-milestone.md

@@ -0,0 +1,43 @@
+---
+description: Start a new milestone cycle — update PROJECT.md and route to requirements
+argument-hint: "[milestone name, for example v1.1 Notifications]"
+---
+
+
+
+## Objective
+Start a new milestone cycle — update PROJECT.md and route to requirements
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse "[milestone name, for example v1.1 Notifications]" from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init new-milestone --raw
+
+3. Load .claude/get-shit-done/workflows/new-milestone.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-new-project.md

@@ -0,0 +1,43 @@
+---
+description: Initialize a new project with deep context gathering and PROJECT.md
+argument-hint: "[--auto]"
+---
+
+
+
+## Objective
+Initialize a new project with deep context gathering and PROJECT.md
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse "[--auto]" from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js init new-project --raw
+
+3. Load .claude/get-shit-done/workflows/new-project.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+

package/.codex/prompts/gsd-pause-work.md

@@ -0,0 +1,43 @@
+---
+description: Create context handoff when pausing work mid-phase
+argument-hint: [none]
+---
+
+
+
+## Objective
+Create context handoff when pausing work mid-phase
+
+## Compatibility
+- Use .codex/skills/get-shit-done-codex semantics.
+- Treat upstream workflow as the source of truth.
+- Replace user-specific paths with workspace-relative paths (.claude/..., .planning/...).
+- Run engine commands through PowerShell: 
+node .claude/get-shit-done/bin/gsd-tools.js ...
+- Parse JSON with ConvertFrom-Json; parse key/value output when workflow uses KEY=value raw mode.
+- No jq / bash-only constructs.
+
+## Subagent lifecycle (required)
+
+- Translate each upstream `Task(...)` into `spawn_agent` -> `wait` -> `close_agent`.
+- Spawn only when the upstream workflow defines an agent role.
+- Use `.claude/agents/gsd-*.md` as role context for each spawned agent.
+- Do not advance workflow steps until wait and close complete.
+## Execution
+1. Parse [none] from the user input.
+2. Run init:
+node .claude/get-shit-done/bin/gsd-tools.js state load --raw
+
+3. Load .claude/get-shit-done/workflows/pause-work.md and execute it step-by-step.
+4. Translate each Task(...) in workflow into:
+   - spawn_agent with the matching role file context from .claude/agents/.
+   - wait for each spawned agent and apply returned output before moving forward.
+5. Preserve all gates and routing from upstream workflow.
+6. Preserve commit behavior using 
+node .claude/get-shit-done/bin/gsd-tools.js commit "message" --files ....
+7. If commit preflight fails (no git / no commit flag), proceed in read-only mode and report clearly.
+
+## Completion output
+- Summarize key artifacts created/updated and next recommended command.
+
+