plan-training-arc

name: plan-training-arc description: "Generate the training plan files for an athlete's arc — primary output is a structured .xlsx spreadsheet (the strategic planning document the athlete actually looks at), with derived markdown files (blocks/, weeks/, active/) for the daily-execution agent to consume. Currently configured for Andy Lee's v6 dunk arc (data hardcoded in build_training_arc.py). Runs ONCE per arc at creation; re-run when arc data changes (load adjustments, exercise swaps, new block design). Trigger when: building a training plan, creating a workout program, generating an arc, or regenerating the training cascade after changes."

plan-training-arc — Skill Guide

Generates the full training cascade for an athlete's arc. Primary output: the .xlsx spreadsheet (outputs/hybrid-athletic-plan-v6.xlsx) — the visual planning document the athlete reviews. Derived outputs: 3 block + 18 week + 2 active markdown files, consumed by the daily-execution agent. Both views render from the same data structures, so they never drift apart.

What this skill does

Runs build_training_arc.py (the implementation). One invocation produces 24 files:

Cadence

Once per arc, at arc creation. Re-run only when arc data changes:

Scope (single-athlete v0)

Currently hardcoded to Andy's v6 dunk arc. The data structures (BLOCK_SPECS, WEEK_META, SQUAT_1RM, BENCH_1RM, CLEAN_1RM) live inline in build_training_arc.py.

When athlete #2 onboards: split data → athlete bundle (e.g., docs/athletes/{name}/arc-{slug}/training/arc-data.py), keep build_training_arc.py as the generic renderer. ~3-hour refactor when needed.

For now: edit the data structures inside build_training_arc.py directly when Andy's plan changes.

How to invoke

python3 .claude/skills/plan-training-arc/build_training_arc.py

Writes to: docs/athletes/andy/arc-2026-summer-dunk/{outputs,training/blocks,training/weeks,training/active}/

When to Use

• User asks to create a training plan, workout program, or periodized program
• User asks to generate an arc, block, or training spreadsheet
• User edited build_training_arc.py data structures and needs to regenerate the cascade
• User wants to formalize goals into a structured training document
• Any request that produces the .xlsx + markdown cascade for a training arc

Required Inputs

Before generating, you need:

1. Athlete profile — name, stats, injuries, schedule, preferences
2. Arc definition — purpose, 2-3 yes/no measurable goals, timeline
3. Block structure — how many blocks, what each block emphasizes
4. Training style(s) — which training-styles/ guides to consult for exercise selection
5. Known maxes — for percentage-based load prescription

If any of these are missing, check docs/athletes/{name}/profile.md and docs/athletes/{name}/arc.md first. If those don't exist, ask the user.

Spreadsheet Structure

Tab 1: Athlete Profile

Tab 2: Arc

Tab 3+: Block Tabs (one per block)

Daily Breakdown Grid Format

# | Exercise | Sets × Reps | Wk 1 | Wk 2 | ... | Wk N

• # column: Exercise numbering with superset notation
  • 1a, 2a, 3a = separate standalone exercises
  • 3a, 3b = superset pair (same number, different letter)
  • 4a, 4b, 4c = tri-set (same number, three letters)
• Loads: Show % + actual weight where maxes are known (e.g., "78% (290 lb)")
• Qualitative loads: "Moderate", "Heavy", "BW", etc. for exercises without tested maxes
• Wave-loaded exercises: Show the actual weight (e.g., "185 lb" → "225 lb" across weeks)

Goal Cascade Model

Every layer follows: 1 purpose → 2-3 yes/no goals

Arc: 1 purpose → G1, G2, G3 (yes/no, test method, deadline)
  └─ Block: 1 purpose → G1, G2, G3 (each with "Serves: Arc G_")
       └─ Week: delivered by text agent
            └─ Day: delivered by text agent

Goals must be:

• Binary — yes or no, not "improve" or "increase"
• Testable — has a specific test method
• Traceable — each block goal explicitly states which arc goal it serves

Exercise Ordering Rules

Within each day, exercises must follow neural demand order:

1. Speed / Plyometrics — when present, always first (max intent required)
2. Power / Olympic lifts — requires coordination and bar speed
3. Strength / Heavy compounds — the primary lift
4. Secondary compounds — moderate load
5. Accessories — superset pairs for efficiency
6. Kick / Sport-specific — technique or power expression
7. Flexibility / Mobility — end of session

Power and Olympic work is NEVER superset. Everything else should be superset where possible.

Color Coding

Style Integration

When building a plan, consult the relevant style guide(s) in docs/content/training-styles/:

• Read the guide's exercise selection, session structure, periodization sections
• Pull exercises from the style's exercise catalog, not generic choices
• Follow the style's ordering rules (e.g., Dylan Shannon: velocity → strength → volume)
• Respect the style's volume and intensity recommendations
• Note exercise substitutions for the athlete's specific injuries

Implementation Notes

• Use openpyxl for spreadsheet generation
• Font: Arial throughout (FT=16pt bold title, FH2=13pt bold section, FB=10pt body)
• Column widths: # col = 6, Exercise = 30, Sets = 15, Week cols = 17
• pct_wt(pct, max) helper: converts percentage to "78% (290 lb)" format, rounds to nearest 5 lb
• Wave-loaded exercises (e.g., 4×5,5,4,3): show actual weight, increase ~10 lb/week
• Deload weeks: reduce volume 60%, intensity to ~75%, maintain all exercises

What This Spreadsheet Is NOT

• Not a daily workout log — the text agent handles that
• Not an exercise library — that lives in docs/content/training-styles/exercises.md
• Not a nutrition plan — that's a separate document
• Not a weekly planner — weeks and days are delivered conversationally

The spreadsheet answers: Where am I, what are my goals, and what's the plan to get there?