The worst fitness plan is usually the perfect one.
Five workouts a week. Carefully split muscle groups. A meal plan. A tracking spreadsheet. A beautiful routine that survives exactly four days.
Then work runs late, the kid gets sick, sleep falls apart, and the plan becomes useless because it only works when life behaves.
That is the problem I want a fitness Agent to solve: not "generate a workout," but repair the plan when reality breaks it.
Start smaller than motivation wants
Most people do not need a harder plan. They need a plan that can survive a bad week.
The open Skill I used for this is the Minimum Viable Fitness Coach Skill. The direct file is SKILL.md.
The Skill is built around a simple idea:
consistency before intensity
That sounds obvious, but most generated workout plans ignore it. They assume the user will do the full version every time.
A better Agent should always include a fallback version.
Full version: 35 minutes
Fallback version: 10 minutes
Emergency version: 5 minutes of easy movement
The fallback is not a failure state. It is part of the plan.
What the Agent should ask before planning
A useful fitness Agent should not jump straight into exercises.
It should ask:
How many days can you realistically train this week?
How much time per session?
What equipment do you have?
Current activity level?
Any pain, injuries, medical restrictions, pregnancy/postpartum status, or red flags?
Main goal: strength, energy, fat loss support, mobility, cardio, habit restart?
What usually causes you to miss workouts?
That last question is easy to skip. It is also the most important one.
If the reason is "I travel," the plan needs hotel-room options. If the reason is "I get sore and quit," the plan needs lower volume. If the reason is "my evenings disappear," the plan needs morning or lunch fallback sessions.
The user's missed workouts are not moral failures. They are data.
The missed-workout repair rule
This is the part I care about most.
A normal plan says:
Monday: strength
Tuesday: cardio
Wednesday: rest
Thursday: strength
Friday: cardio
Then the user misses Monday and Tuesday.
A bad Agent says: "No worries, get back on track," which is nice but not useful.
A better Agent repairs the week:
You missed two sessions. Do not try to cram all four workouts into the next three days.
This week, do:
- one short full-body strength session
- one easy walk
- one optional 10-minute mobility session
Then review what blocked Monday and Tuesday.
The Skill treats missed workouts as an input to the next plan. That is the difference.
RPE beats fake precision
For beginners and people restarting, exact percentages are often overkill.
The Skill uses RPE, or rate of perceived exertion. In plain English: how hard the set felt.
Most beginner working sets should sit around:
RPE 6-8
That means challenging, but not a grind. The user should usually have a few reps left in reserve.
This helps the Agent avoid two bad plans:
too easy to matter
too hard to repeat
Progress should be small and boring:
add one set
add a few reps
add a little load
add a few minutes
Not all at once.
Weekly review is where the plan improves
The Agent should not progress the plan just because seven days passed.
It should ask what actually happened.
How many sessions did you complete?
What was average RPE?
Any pain?
What got in the way?
Which session felt easiest to repeat?
Which one did you avoid?
Then apply simple rules:
Completion below 60%: reduce the plan and remove friction.
Completion 60-80%: repeat the plan and improve scheduling.
Completion above 80% with manageable RPE: progress slightly.
Completion above 80% but RPE too high: repeat or reduce intensity.
Pain or red flags: modify conservatively and consider medical help.
This is more useful than a motivational speech.
The safety boundary matters
Fitness advice can drift into medical advice quickly.
The Agent should not diagnose injuries, write rehab plans, override a clinician, or tell someone to push through significant pain.
Before creating or intensifying a plan, it should screen for red flags like:
chest pain or pressure
fainting or dizziness
unexplained shortness of breath
pain that worsens with movement
clinician-imposed restrictions
major heart, lung, metabolic, neurological, or surgical concerns
If those show up, the Agent should back off and recommend medical evaluation instead of generating a harder workout.
A fitness Agent should be useful. It should not be reckless.
What to remember
For this kind of Agent, memory should not be generic.
It should remember the user's actual constraints:
available days
time per session
equipment
current baseline
injury cautions
energy patterns
missed-workout patterns
preferred exercises
exercises they hate
weekly completion history
RPE notes
That memory is what makes the next plan better than the first one.
If the user keeps missing evening workouts, the Agent should stop suggesting evening workouts. If push-ups irritate the wrist, the Agent should remember the substitution. If 20 minutes works and 40 minutes fails, the next plan should respect that.
Using the Skill in your own setup
If your agent client can install, import, or reference GitHub-based Skills, use the Minimum Viable Fitness Coach Skill as the source.
For a repo-based agent project, keep it as:
skills/minimum-viable-fitness-coach/SKILL.md
Then load it when the user asks about starting fitness, restarting after a break, missed workouts, beginner strength, walking/cardio, mobility, or weekly review.
For Claude Code or another repo-aware coding agent, the Skill can guide your agent instructions, safety screen, plan templates, and state schema.
For ChatGPT-style custom assistants, use the Skill as the source of truth if your setup supports imported instructions or external Skill references.
The Skill is the coaching method. The user's schedule, constraints, and weekly history are memory.
Start using it in Telegram or WhatsApp
If you want to try this without setting up the Skill yourself, you can create the Minimum Viable Fitness Coach Agent on ClawMama and use it directly in Telegram or WhatsApp.
The Skill is already attached, so you can focus on building the habit instead of maintaining prompts. The Agent can keep track of your schedule, equipment, missed-workout patterns, energy level, and weekly reviews.
A useful test
Ask the Agent this:
I planned to work out four times this week, but I already missed the first two days. What should I do now?
A weak Agent says to stay motivated.
A better Agent reduces the plan, protects recovery, asks what got in the way, and gives you a version you can still complete this week.
That is the fitness Agent I would rather use.
Top comments (0)