Skip to content

nikitavovch/claude-gym-coach / skills/exercise

coach-exercise

Verified

This skill should be used when the athlete wants the coach to know an exercise it doesn't recognize, or another name for one it does — "add an exercise", "добавь упражнение", "у меня своё упражнение", "the coach doesn't know X", "я называю это по-другому", "запомни, что X — это Y", or runs /coach:exercise. It adds a row to the data folder's exercises.md, or an alias to an existing row, through the statistics script.

Coach Exercise

Teaches the folder's exercise table something new: a whole exercise the table doesn't have, or another name for one it does. Everything else in this plugin resolves exercise names through that table — log writes canonical names by it, report groups volume by its muscles, exercises and every record are keyed by it — so a row added carelessly doesn't just sit there: a new name that some existing exercise already answers to takes that name over, and every session ever logged under it silently moves to the new row. That is why this skill never edits exercises.md itself. It gathers the answers and hands them to stats.mjs exercise add or stats.mjs exercise alias, which refuse any name already taken, any muscle or equipment the table doesn't use, and any | or ; inside a name, and only then write the file.

This document is written in English, for the model. Everything the coach says to the athlete during a run is in the athlete's own language.

Plugin files — the statistics script, anything under knowledge/ or agents/ — are addressed by their full path, which Claude Code fills in before this text is read. The athlete's own files (.coach.json, exercises.md, athletes/<id>/...) are relative to the working directory, which during a run is the athlete's data folder, not the plugin.

1. Guard

Check for .coach.json in the working directory. Missing — tell the athlete to run /coach:init and stop. Present — check its schema field: anything other than 1 means the plugin and the data folder are out of sync, so say so and stop without attempting to auto-migrate.

exercises.md is shared by everyone in the folder, so there is no athlete to resolve — but say so if two athletes share the folder: a row added here is added for both.

2. Is it already known?

Take the name from the argument, or ask for it — exactly as the athlete says it. Then run:

node "${CLAUDE_PLUGIN_ROOT}/scripts/stats.mjs" exercises --stdin --json <<'COACH_JSON'
{"names": ["<the name exactly as the athlete said it>"]}
COACH_JSON

known: true means the table already answers to that name: tell the athlete which exercise it resolves to (canonical) and stop. Nothing to add.

3. Another name, or a new exercise?

Most unknown names are a known exercise said differently — "тяга Мидса" for the table's Тяга Мидоуза, "присед Андерсона" for Присед от пинов. Read exercises.md and look for rows that plausibly mean the same movement: same implement, same pattern, a shared word in the name or aliases. Also check ${CLAUDE_PLUGIN_ROOT}/knowledge/exercises.md §9 — a bare "жим", "тяга", "разводка" and the like are ambiguous on purpose, and the answer to those is the qualifying word, never a new row.

  • One plausible match: ask a single yes/no question — "это Присед от пинов?". On yes, go to §4 with that row.
  • Several: list them and ask which one, with "none of these — a new exercise" as the last choice.
  • None, or the athlete says it's something different: go to §5.

Never decide this silently. Merging two different movements hides the second one's history inside the first; splitting one movement in two halves its records.

4. Add the name to an existing exercise

node "${CLAUDE_PLUGIN_ROOT}/scripts/stats.mjs" exercise alias --stdin <<'COACH_JSON'
{"name": "<existing name>", "aliases": ["<new name>", "<another>"]}
COACH_JSON

name can be the canonical name or any alias it already has. Add every spelling the athlete actually uses, in one call. The athlete's words go in the JSON exactly as they said them: the quoted 'COACH_JSON' delimiter stops the shell from expanding anything between the lines, so the only escaping is JSON's own (\" for a quote inside a name). Go to §7.

5. Add a new exercise

Gather, in one message rather than a question at a time:

  • The name in the athlete's language, as they'd want to read it in a plan — this becomes the canonical name, and it is permanent: logs are written with it, so it is never renamed later.
  • Other names they use for it — the dictated spelling always goes in, if it isn't the canonical name itself.
  • Equipment, using the table's own tags (barbell, dumbbell, bench, cable, bodyweight, … — the full list is in ${CLAUDE_PLUGIN_ROOT}/knowledge/exercises.md §3), or machine:<name> for a specific machine. Check it against the athlete's gym inventory; an exercise their gym can't do is worth a word, but it is theirs to add.
  • Muscles, primary first — report counts weekly sets under the first one — from the same §3 vocabulary.
  • Main lift or not. yes only for a heavy multi-joint movement they load progressively and want trends and stall checks on (the criterion in knowledge/exercises.md §4); validate then insists on RPE for it. When in doubt, no.

Show the row back in plain words and wait for a yes. Then:

node "${CLAUDE_PLUGIN_ROOT}/scripts/stats.mjs" exercise add --stdin <<'COACH_JSON'
{"name": "<name>", "aliases": ["<a>", "<b>"], "equipment": ["<e1>", "<e2>"], "muscles": ["<primary>", "<secondary>"], "main": false}
COACH_JSON

Same heredoc rule as §4: the athlete's text goes into the JSON as said.

6. When the script refuses

Every refusal names what to fix; relay it and ask, never work around it:

  • already names <exercise> — that name belongs to another exercise. Either the athlete meant that exercise after all (go back to §3), or the new one needs a different name or a qualifier.
  • unknown muscle / unknown equipment — the message lists the table's vocabulary; pick from it with the athlete. A new muscle name would split their weekly volume by a spelling.
  • holds "|" or holds ";" — those characters separate the table's columns and lists; rephrase without them.
  • which a shell would act on — a name with ", a backtick, $ or \ in it would run as code on a later command line; rephrase without it.

7. Reply

Say what the table now knows, in one or two lines, from the script's output. Mention that every earlier log that used the new name now counts toward that exercise — names are resolved when a log is read, not when it was written, so nothing needs re-logging.

Rules that apply throughout

  • exercises.md is written only through stats.mjs exercise add and stats.mjs exercise alias, with --stdin — never with an editor, never by hand, never with the athlete's words on the command line.
  • Nothing is renamed or removed. A canonical name is what the logs hold; a spelling the athlete dislikes gets a new alias, not a rename.
  • Merging or splitting movements is always the athlete's call (§3).
  • Plugin-side text (this file) is English; everything said to the athlete, and every name written for them, is in their language.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall