agentleFS
kaustin923kaustin923/agent-fitness-coach/.claude/skills/plan-adjust/SKILL.md

plan-adjust

Skill0 starsChanged 4 months agoMITView on GitHub

Modify the active training plan safely — single-day edits or plan-wide restructures (reschedule, volume up/down, swaps, frequency, priorities, missed-week recovery) with invariant re-checks and file-verified narration. Use when the athlete wants to move, swap, or change a workout, make the plan easier or harder, change which days they train, recover from missed sessions or illness, or when weekly-review recommends an adjustment.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Prefer adjusting over rebuilding.

    line 10
  2. Very specific:

    Read the full future scope: every day from today through meta.end_date.

    line 58
  3. Distinctive:

    All date math, mileage scaling, and sums run in python.

    line 91
  4. Very specific:

    If the athlete reports pain or injury as the reason for adjusting, adjust conservatively *and* apply the injury rules in …

    line 93
  5. Distinctive:

    Only claim "adjusted" or "moved" after the file write happened this turn.

    line 94

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Plan adjust

A plan that can't bend gets abandoned. Your job is to bend it without breaking it: change exactly what needs changing, keep every structural guarantee intact, and — the cardinal rule — describe the result from what the file actually says after the write, never from what you intended to write.

Prefer adjusting over rebuilding. Regenerate from scratch (plan-builder) only when the goal or race itself changes; everything else is an adjustment.

When to run

  • The athlete asks directly: "move Thursday's tempo to Saturday", "this week is too much", "I was sick all week", "make the plan harder", "swap Wednesday for a bike ride", "I can only run 3 days now".
  • weekly-review produced an adjustment suggestion (completion < 60% → reduce frequency; the same weekday repeatedly skipped → move that workout; skips citing fatigue → reduce volume; ≥ 80% completion → leave the plan alone).
  • The athlete has missed 3+ consecutive scheduled training days (rest and cross-train-only days count neither way): offer a reconfigure once, and if declined, don't offer again for 7 days.

Inputs

  • athlete/plan.json — the canonical plan. No active plan → offer plan-builder instead.
  • athlete/profile.md — schedule constraints and declared units.
  • athlete/log.json — completion and actuals context (what's driving the adjustment).
  • Reference: reference/training-rules.md ("Plan adjustment rules", "The whole-mile rule and other invariants", "Strength programming"), reference/safety.md.

Procedure

1. Dates first

Compute today with date +%F and build the weekday↔date table for this week and next (the COACH.md ritual). Every day you touch is identified by its date and weekday, both read from the table. "Next Monday" computed in your head is how the wrong workout gets moved.

2. Choose the tier

Tier 1 — single-day edit: the request names one day and stays on it. Change a workout's date, distance, duration, type, or details. Apply directly.

Tier 2 — plan-wide adjustment: anything touching multiple days or weeks. Classify it before acting:

TypeMeaning
reschedulemove workouts to different days, content intact
reduce_volumescale run miles or lift sets/duration down by a stated percent
increase_volumescale up, within progression limits
swap_workoutreplace one session type with another
skip_and_adaptabsorb missed sessions; redistribute what matters, drop what doesn't
reprioritizeshift emphasis between running and lifting
change_frequencychange training days per week, every remaining week
customanything else — state your interpretation before applying

3. Tier 1 — single-day edit

  1. Locate the day in plan.json by date. If the named day holds only a Rest day, correct the premise instead of silently moving the rest row: "Thursday's actually a rest day — your next session is Friday's tempo. Want me to move that?" If multiple candidates match a fuzzy description, ask one clarifying question.
  2. Edit that one day object in python, respecting constraints 8–12 below. Moving a day within its week: the content travels intact; update date and weekday together, and re-sequence the week's days so it stays Mon–Sun ordered.
  3. Re-validate the affected week: segment sums and weekly_miles = day sum always; whole-mile totals and the 48 h lower-body/hard-run spacing as the athlete has them set — a deviation they chose is recorded, not reverted.
  4. Re-render athlete/PLAN.md from plan.json (projection conventions per templates/PLAN.example.md).
  5. Re-read the changed week from the file and confirm from it: "Done — tempo is now Saturday 07-11, Thursday 07-09 is rest."

4. Tier 2 — plan-wide adjustment

  1. Read the full future scope: every day from today through meta.end_date. Days in the past and days with status completed or skipped are read-only history — adjustments apply today-forward only, and history is context for the redesign, never a casualty of it.
  2. Draft the modification set before touching the file: a one-line summary plus explicit lists of updates (day, field, old → new), deletes, and inserts. Every insert that replaces a day pairs with a delete of what was there. If the request is ambiguous in a way that changes the outcome materially ("reduce volume" — by how much?), ask one question; otherwise state your interpretation in the summary and proceed.
  3. Apply in one python pass. Recompute every touched week's weekly_miles and re-title/re-describe any day whose content changed.
  4. Re-run the full invariant checklist from the plan-builder skill — arithmetic and structural integrity (segment sums, week sums, weekday↔date agreement, rest-day preferences) always; preference and methodology rules (whole miles, progression limits, 48 h spacing) as the athlete has them set. A failed arithmetic check means fix and re-run — never leave plan.json arithmetically broken, and never present it that way. A preference-rule deviation the athlete chose gets recorded in meta.methodology, not silently reverted.
  5. Update meta: days_per_week for frequency changes; user_preferences for new firm constraints or priorities; append a dated one-liner to meta.methodology ("2026-07-06: cut W6–W8 volume 20% — illness recovery").
  6. Re-render athlete/PLAN.md.
  7. Verify, then narrate: re-read plan.json and describe the next two weeks only from what the file now contains. If you deliberately limited scope (say, only restructured through week 10 of 16), say so explicitly rather than implying the whole plan changed.

5. The adjustment constraints

These thirteen rules are the contract (canonical statement in reference/training-rules.md, "Plan adjustment rules"):

  1. Minimize disruption — modify only what the request requires; every untouched day stays byte-identical.
  2. Keep the plan's structure — phases, cutback weeks, and taper survive unless they are the target of the request.
  3. Reschedules move days intact — type, title, prescription, and strength cue travel with the date.
  4. Volume changes scale honestly — cut or grow run miles, or lift sets/duration, by the stated percent, then re-fit sessions to the whole-mile rule and recompute weekly totals.
  5. Reprioritizing reorders emphasis, not identity — protect the prioritized modality's key sessions, trim the other's, and record the new priority in meta.user_preferences.
  6. skip_and_adapt redistributes, never crams — carry forward the most important missed work (long run first, then quality) without overloading any single day or pushing a week past progression limits; let the rest go.
  7. Frequency changes apply to every remaining week — not just the next one.
  8. Never schedule heavy lower-body lifting within 48 hours of a hard run (Tempo, Interval, Long, Race) — in either order.
  9. Edit days in place for field changes — a moved or resized workout keeps its day-object identity; wholesale replacement is only for discipline changes.
  10. Retyping to a run or lift demands complete content — a day becoming a run needs workout, pace_zone, and mileage fields; a day becoming a lift needs duration_min and lift_details with real exercises. A hollow retype is invalid. Retyping in place is only safe toward Rest or Cross-Train.
  11. Rest days are clean — no miles, duration_min: null, and stale run/lift fields stripped when a day becomes Rest.
  12. No orphaned days — every scheduled non-rest day ends the edit with a prescription and a duration or mileage; every replacing insert pairs with a delete.
  13. Summarize clearly — the athlete hears what changed, what didn't, and why, in plain language.

These constraints discipline the coach, not the athlete. If the athlete explicitly asks for something a training guardrail here would block — say, heavy legs the day before their long run — give one concise, specific risk note; if they confirm, make the change their way, record it in meta.user_preferences, and don't re-warn on later edits. (The data-integrity constraints — 9 through 12 — aren't about athlete choices and always hold.)

Rules

  • Narrate the file, not your intent. "The plan now shows…" must be literally true of athlete/plan.json at the moment you say it. The single worst failure mode of this skill's ancestors was announcing schedule changes that were never written.
  • Completed and skipped days are immutable history. So are their actual blocks.
  • plan.json stays canonical; PLAN.md and the dashboard are regenerated projections. Offer to refresh the dashboard after any adjustment.
  • All date math, mileage scaling, and sums run in python.
  • Destructive requests — archiving or abandoning the plan — get a one-line confirmation first ("Archive your marathon plan? It has 6 weeks left."), then set meta.status: "archived". See reference/safety.md.
  • If the athlete reports pain or injury as the reason for adjusting, adjust conservatively and apply the injury rules in reference/safety.md — reduced volume is not a treatment plan.
  • Only claim "adjusted" or "moved" after the file write happened this turn.

Output

Show: the one-line summary, a compact diff of changed days (date · was → now), the verified next-two-weeks schedule read back from the file, and the invariant-check result.

Then suggest 2–3 next actions (skip if you just asked a question): regenerate the dashboard (dashboard), sync activities so completions stay current (strava-sync), or a morning daily-checkin to walk the new week.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.