feat(workout): set UX overhaul -- checkbox confirm, swipe delete, gym picker fix, rest minimize

.claude/commands/adr.md

@@ -1,6 +1,7 @@
 ---
 description: Create or review Architecture Decision Records
 model: sonnet
+effort: high
 ---
 
 # Architecture Decision Record

.claude/commands/backlog.md

@@ -1,8 +1,9 @@
----
-description: Add to or prioritize the backlog
-model: sonnet
----
-
+---
+description: Add to or prioritize the backlog
+model: sonnet
+effort: low
+---
+
 # Backlog Management
 
 Add ideas or bugs to the backlog, or prioritize existing items. Keeps track of future work and known issues.

.claude/commands/blueprint.md

@@ -1,6 +1,7 @@
 ---
 description: Full 4-phase planning: Spec, Tech, Steps with team orchestration
 model: sonnet
+effort: high
 ---
 
 # Full Planning Workflow

.claude/commands/commit.md

@@ -1,8 +1,9 @@
----
-description: Git commit with conventional format and task references
-model: sonnet
----
-
+---
+description: Git commit with conventional format and task references
+model: sonnet
+effort: low
+---
+
 # Commit Changes
 
 Prepare a commit message following conventional commits format. References tasks and ADRs from planning artifacts.
@@ -56,3 +57,10 @@ After commit is created:
   - Start testing
   - Update documentation
   - Review and verify
+
+### Step 4: Archive completed QuickPlans
+After a successful commit, check for active QuickPlans:
+1. Scan `Context/QuickPlans/` for `.md` files (ignore `Done/` subdirectory)
+2. If none found, skip this step silently
+3. If plans are found, list them and ask: "Archive these completed QuickPlans?"
+4. If confirmed, move each plan to `Context/QuickPlans/Done/` (create the directory if needed)

.claude/commands/diagnose.md

@@ -1,6 +1,7 @@
 ---
 description: Check project structure and Cortex health
 model: sonnet
+effort: low
 ---
 
 # Diagnose: Cortex Project Health Check

.claude/commands/impl.md

@@ -1,6 +1,7 @@
 ---
 description: Execute Steps.md via hub-and-spoke sub-agent orchestration
 model: sonnet
+effort: medium
 ---
 
 # Implementation Orchestrator
@@ -21,14 +22,14 @@ Execute a planned feature by reading Steps.md and delegating tasks to specialist
 3. Once identified, load Steps.md
 
 ### Step 2: Load planning context
-1. Read `Context/Features/NNN-FeatureName/Steps.md` -- tasks, team, milestones
-2. Read `Context/Features/NNN-FeatureName/Spec.md` -- requirements and assertions
-3. Read `Context/Features/NNN-FeatureName/Tech.md` -- architecture decisions
-4. Read `CLAUDE.md` for project conventions
-5. Parse Steps.md to extract:
+1. Read `Context/Features/NNN-FeatureName/Steps.md` -- this is the ONLY planning file the orchestrator reads
+2. Parse Steps.md to extract:
    - Team Members (names, agent types, resume behavior)
    - Task list with assignments, dependencies, parallel flags
    - Milestones with testable assertion references and contract declarations
+   - File ownership boundaries per agent
+
+**Context budget rule:** Do NOT read Spec.md, Tech.md, CLAUDE.md, or any source code into the orchestrator's context. Agents read these files themselves. The orchestrator's job is coordination, not comprehension -- if it compacts before wave 1 starts, the entire execution plan is lost.
 
 ### Step 3: Initialize execution session
 1. Create `.cortex/session.md` with plan reference and team roster
@@ -147,14 +148,16 @@ After all waves complete:
 
 ## Agent Prompt Template
 
-When rendering each agent's prompt file, use this template:
+When rendering each agent's prompt file, use this template. The orchestrator fills in task IDs, file paths, and milestone references from Steps.md -- it does NOT read Spec.md/Tech.md to paste content. Agents read those files themselves.
 
 ```
 You are [NAME], the [ROLE] on this team.
 
 FEATURE CONTEXT:
-[Paste relevant Spec.md sections -- requirements and testable assertions]
-[Paste relevant Tech.md sections -- architecture decisions for this domain]
+Read these files before starting work (focus on sections relevant to your tasks):
+- Context/Features/NNN-FeatureName/Spec.md -- requirements and testable assertions
+- Context/Features/NNN-FeatureName/Tech.md -- architecture decisions
+- CLAUDE.md -- project conventions
 
 YOUR TASKS:
 [List specific tasks from Steps.md assigned to this agent]
@@ -182,7 +185,6 @@ COORDINATION (event log):
 FILE OWNERSHIP RULES:
 - Mark your tasks complete via TaskUpdate when done
 - If you change an interface another agent depends on, note it clearly in `.cortex/session.md`
-- Read all relevant files and load skills before starting work
 
 Think hard and provide thorough implementation.
 ```
@@ -301,3 +303,5 @@ When to resume vs start fresh:
 - Contract files must exist before downstream agents are spawned
 - The quality-engineer always runs last in validation mode (read-only)
 - Advisor-cli exit code 2 means fall back to in-thread reasoning -- never fail the wave on advisor unavailability
+- **The orchestrator must NOT read source code, explore the codebase, or load files beyond Steps.md.** Codebase comprehension is the agents' job. The orchestrator reads Steps.md, builds prompt files, spawns agents, and monitors events. Any additional file reads risk compaction before wave 1.
+- **No pre-flight advisor consult.** The advisor is called at milestone boundaries (Step 5e), not before wave 1. Pre-flight exploration and advisor calls are the two largest sources of unnecessary orchestrator context.

.claude/commands/qa.md

@@ -1,8 +1,9 @@
----
-description: Run stack-specific QA agents in parallel
-model: sonnet
----
-
+---
+description: Run stack-specific QA agents in parallel
+model: sonnet
+effort: medium
+---
+
 # Parallel Quality Assurance
 
 Run stack-specific QA agents in parallel across all detected stacks in the project.

.claude/commands/quick.md

@@ -1,6 +1,7 @@
 ---
 description: Lightweight single-file plan for small tasks
 model: sonnet
+effort: medium
 ---
 
 # Quick Plan
@@ -53,5 +54,10 @@ ALWAYS end by recommending the most appropriate execution command:
 
 Quick plans typically use `/impl` since they are smaller in scope.
 
-### Step 5: Clean up
-When task is complete, move quick plan to `Context/Backlog/Done/` for reference, or delete if not needed.
+### Step 5: Archive the QuickPlan
+When the task is complete, archive the plan so it does not accumulate in the active folder:
+1. Create `Context/QuickPlans/Done/` if it does not exist
+2. Move the plan file from `Context/QuickPlans/` to `Context/QuickPlans/Done/`
+3. Confirm the move to the user
+
+Do NOT skip this step or leave it as a suggestion. Perform the move directly.

.claude/commands/review-capture.md

@@ -1,8 +1,9 @@
----
-description: Capture review findings with 4-category triage
-model: sonnet
----
-
+---
+description: Capture review findings with 4-category triage
+model: sonnet
+effort: medium
+---
+
 # Review Capture
 
 Capture review feedback and link it to the feature or PR being reviewed. Documents what was reviewed, feedback provided, and any decisions made.

.claude/commands/review-resolve.md

@@ -1,6 +1,7 @@
 ---
 description: Work through captured review findings with interactive triage
 model: sonnet
+effort: high
 ---
 
 # Review Resolution

.claude/commands/review-verify.md

@@ -1,6 +1,7 @@
 ---
 description: Verify all review findings have been addressed
 model: sonnet
+effort: medium
 ---
 
 # Review Verification

.claude/commands/security-scan.md

@@ -1,6 +1,7 @@
 ---
 description: Scan Cortex and Claude Code configuration for security issues
 model: sonnet
+effort: medium
 ---
 
 # Security Scan: Configuration Audit

.claude/commands/setup.md

@@ -1,6 +1,7 @@
 ---
 description: Initialize project for Cortex workflow
 model: sonnet
+effort: low
 ---
 
 # Setup Cortex

.claude/commands/spec-init.md

@@ -0,0 +1,117 @@
+---
+description: Initialize spec-driven design templates for the project
+model: sonnet
+effort: low
+---
+
+# Initialize Spec Templates
+
+Generate the Spec-Driven Design template set for this project. Scans existing code and documentation, then produces structured spec docs in `docs/specs/` that integrate with the Cortex planning workflow.
+
+## When to use
+- First time setting up spec-driven development in a project
+- Converting existing docs (PRDs, architecture docs, Cortex specs) into the template format
+- User says "init specs", "create spec templates", "set up specs", "convert my docs"
+
+## Workflow
+
+**CRITICAL:** This command reads project context before generating anything. Do NOT write files until Step 4.
+
+### Step 1: Gather project context
+1. Read `CLAUDE.md` for project name, tech stack, architecture, conventions
+2. Scan project structure for existing documentation:
+   - `docs/` -- any existing spec or design docs
+   - `README.md` -- project description
+   - `Context/Features/` -- existing Cortex feature specs (Spec.md, Tech.md, Steps.md)
+   - `Context/Decisions/` -- existing ADRs
+3. Scan source code structure for domain hints:
+   - Model/entity class definitions
+   - Enum/sealed class definitions
+   - Database migration files
+   - Route/endpoint definitions
+4. Check if `docs/specs/` already exists
+
+### Step 2: Present findings and choose mode
+Present what was found and offer three modes:
+
+```
+Project: {name}
+Stacks: {detected stacks}
+Existing docs found: {list}
+Cortex specs found: {count} features in Context/Features/
+
+Modes:
+  [G] Greenfield -- blank templates pre-filled from project context
+  [B] Backfill -- convert existing docs into template format
+  [S] Selective -- pick which templates to generate
+
+Which mode? [G/B/S]:
+```
+
+Wait for user choice before proceeding.
+
+### Step 3: Determine template set
+For greenfield and backfill, confirm which templates to generate:
+
+**Always generate (singletons):**
+- `00-project-overview.md`
+- `02-domain-model.md`
+- `03-invariants.md`
+- `04-architecture.md`
+- `05-erd.md`
+- `06-state-machines.md`
+- `07-user-flows.md`
+
+**Ask the user (multi-instance):**
+- How many PRDs? What are they named? (suggest based on existing docs/features)
+- Any subsystem deep-dives needed? (suggest: notifications, auth, sync if detected)
+- Any secondary platform docs? (suggest based on detected platforms)
+- How many implementation plans? (suggest: one per major feature area)
+
+### Step 4: Generate templates
+Use the `spec-init` skill (see `core/skills/spec-init/SKILL.md`) to generate each template file.
+
+For each file:
+1. Create the file in `docs/specs/` with correct naming convention
+2. In greenfield mode: fill project-specific placeholders, keep guidance comments
+3. In backfill mode: map existing doc content to template sections, flag inferred content
+4. Show progress as files are generated
+
+### Step 5: Validate and report
+1. Run cross-reference validation (assertions, invariants, entities, states)
+2. Generate `docs/specs/GAPS.md` with sections needing attention
+3. Present summary:
+
+```
+Spec Init Complete
+  Mode: {greenfield/backfill/selective}
+  Files generated: {count}
+    - 00-project-overview.md
+    - 01-PRD-001-core.md
+    - 01-PRD-002-sync.md
+    - 02-domain-model.md
+    - ...
+
+  Cross-references:
+    [PASS] 12/12 P0 requirements have assertions
+    [WARN] 2 invariants need assertion mappings
+
+  Gaps: docs/specs/GAPS.md (3 sections need attention)
+
+Next steps:
+  1. Review GAPS.md and fill empty sections
+  2. Run /blueprint for your first feature
+  3. plan-from-specs will auto-detect these docs
+```
+
+## Rules
+- Always run Step 1 (context gathering) before generating anything
+- Never overwrite existing files in `docs/specs/` without asking
+- In backfill mode, preserve original docs -- copy content, don't move/delete
+- Show progress as files are generated (don't go silent for 30 seconds)
+- If the project has no CLAUDE.md, suggest running `/setup` first
+- Cross-reference validation is not optional -- always run it
+- GAPS.md is always generated, even when no gaps are found
+- For backfill mode, prefer doc content over source code inference
+- Flag all inferred content with `<!-- INFERRED: verify this -->` comments
+- If `docs/specs/` already has files, ask before proceeding (merge vs overwrite vs abort)

.claude/commands/team-impl.md

@@ -1,6 +1,7 @@
 ---
 description: Execute Steps.md via Agent Teams with peer-to-peer coordination
 model: sonnet
+effort: medium
 ---
 
 # Team Implementation (Agent Teams)