readiness-check

name	readiness-check
description	Pre-run safety gate that evaluates readiness and recommends proceed/modify/skip decisions. Use before scheduled workouts to assess recovery status, injury signals, and training load based on sleep, soreness, and recent activity.
metadata	{"short-description":"Pre-run safety and readiness gate with conservative recommendations.","agent":"cursor"}

When Cursor should use this skill

• Before starting a planned workout or recording a run
• When the user reports fatigue, soreness, or poor sleep
• When user asks if they should run today or if they're ready for a workout
• When implementing pre-run readiness features or debugging readiness assessments

Invocation guidance

1. Supply UserProfile, recent TrainingHistory, and selfReport (sleep, soreness, mood).
2. Evaluate against load/monotony caps and health signals; prefer conservative outcomes.
3. Return a readiness decision with SafetyFlag[] and recommended modifications.
4. Calculate training load from past 7-14 days.
5. Check for injury signals, illness, or extreme fatigue.

Input schema

See references/input-schema.json.

Output schema

See references/output-schema.json.

Decision framework

PROCEED

• Good sleep (7+ hours)
• Low/moderate soreness
• No pain or illness
• Training load within normal range
• Good mood/energy

MODIFY

• Poor sleep (5-7 hours)
• Moderate soreness
• Minor fatigue
• Slight load spike (15-25%)
• Moderate stress Modifications: Reduce intensity OR volume by 20-30%, convert hard → easy

SKIP

• Very poor sleep (<5 hours)
• High soreness or pain
• Illness symptoms
• Extreme fatigue
• Major load spike (>30%)
• Injury signals Recommendation: Rest day, consider cross-training if feeling better later

Integration points

• UI:
  • Pre-run modal (before GPS start)
  • Today screen readiness widget
  • Disable GPS start if decision is skip with override option
• API: v0/app/api/run/readiness (to add)
• Background: Can run nightly to precompute next-day readiness
• Database: Store readiness checks in readiness_checks table (if added)

Safety & guardrails

• If pain/dizziness/injury keywords detected → decision must be skip, advise stop and consult professional.
• If data missing or uncertain → default to modify, emit SafetyFlag missing_data.
• Never provide medical diagnosis.
• Err on side of caution - when in doubt, recommend modify or skip.
• If user overrides skip decision, log override and suggest extra caution.

Self-report indicators

Sleep quality

• Good: 7-9 hours, restful
• Fair: 5-7 hours or restless
• Poor: <5 hours or very restless

Soreness level

• None: No soreness
• Mild: Some stiffness, goes away with movement
• Moderate: Noticeable soreness, affects movement
• Severe: Pain that alters gait

Mood/Energy

• Energized: Eager to run
• Normal: Ready to run
• Fatigued: Low energy
• Exhausted: Drained, struggling

Stress level

• Low: Relaxed, calm
• Moderate: Some stress, manageable
• High: Very stressed, overwhelmed

Training load assessment

Acute:Chronic Workload Ratio (ACWR)

• Safe: 0.8-1.3
• Caution: 1.3-1.5 or <0.8
• Danger: >1.5

Week-over-week volume change

• Safe: <20% increase
• Caution: 20-30% increase
• Danger: >30% increase

Consecutive hard days

• Safe: 0-1 hard days in past 3 days
• Caution: 2 hard days in past 3 days
• Danger: 3+ hard days in past 3 days

Telemetry

• Emit ai_skill_invoked and ai_safety_flag_raised (if any) with:
  • decision (proceed/modify/skip)
  • safety_flags
  • model
  • latency_ms
  • override_count (if user overrides)
  • factors (array of contributing factors)

Common edge cases

• First run ever: Bias toward proceed with encouragement
• Return from injury: Bias toward modify, very conservative
• After illness: Skip if any symptoms, modify if recovered <7 days
• Race day: Allow proceed even with mild fatigue (but warn)
• Taper period: Bias toward proceed (reduce pre-race anxiety)
• Missing data: Request more info, default to modify

Testing considerations

• Test all three decision outcomes (proceed/modify/skip)
• Verify SafetyFlag emission for each danger signal
• Test with missing self-report data
• Validate modification suggestions match fitness level
• Test load calculations accuracy
• Verify override behavior and logging