gtapps/domain-brainstorm

Domain Brainstorm

Kill criteria (read before running)

After ≥8 invocations, read .claude-code-hermit/state/proposal-metrics.jsonl and filter events where type:"brainstorm-emit" and skill:"domain-brainstorm". Count CREATE vs total emits with ideas (triage-survival) and read status: of the resulting PROP files (PROP-acceptance). If triage-survival < 25% or PROP-acceptance < 30%, cut this skill rather than tune it — signal-to-noise isn't there.

Gate 0 — Gather inputs

Start with the connection check; then read the remaining sources in parallel. Do not run tests or install tools.

Connection check Call mcp__strava__check-strava-connection. If disconnected, stop immediately:

Domain brainstorm aborted — Strava disconnected. Reconnect and retry.

Stated goals Read MEMORY.md (the index). For each entry whose title or description suggests goals, training preferences, or athlete profile, read that topic file. Extract stated goals and named disciplines (e.g. "marathon training", "strength 2×/week", "sleep 8h target"). If no goals are parseable from MEMORY.md, derive implicit goals from the Strava activity mix gathered below — note in every grounding citation that the goal is inferred, not stated.

What's been logged Delegate bulk aggregation to @claude-code-fitness-hermit:strava-data-cruncher:

Last 8 weeks of activities. Return: per-week count + total minutes by sport/activity type. Flag any sport with zero sessions in the last 4 weeks.

This keeps heavy multi-page fetching off the main context and respects rate limits.

Logged subjective and coverage signals Glob .claude-code-hermit/compiled/activity-*.md and any .claude-code-hermit/compiled/weekly-*.md files modified within the last 30 days — read the first 15 lines of each (frontmatter + opening paragraph). Read state/activity-notes.json to check which signal dimensions (RPE, sleep, macros) actually have entries vs gaps.

Gate 1 — Generate ideas (max 2)

Think across all inputs simultaneously. For each candidate, both constraints must pass before including it:

1. Concrete gap — state the coverage or imbalance in one sentence: what goal or discipline has no recent supporting data, or what mix is visibly wrong vs stated goals? If no specific gap can be named, discard.
2. ≥2 named grounding items — cite at least two by name (e.g. memory:athlete-profile, strava:8wk strength=0sessions, notes:sleep last=21d ago, compiled:activity-2026-05-12). These support the gap; the gap is the bar.

Map each passing idea to the closest fitness prefix:

• [goal-gap] — a stated or inferred goal with no recent supporting data. Covers: a discipline not logged in N weeks, sleep or macro dimensions the operator tracks but hasn't entered recently, a race/event goal with no recent specificity work.
• [imbalance] — training mix wrong vs stated or inferred goals. Covers: cardio/strength ratio skew, the same workout type repeated across multiple weeks with no progression toward a stated goal.

Scoping guard: never emit ideas about cardiac-drift trends, HR/pace efficiency slope, recovery-score trends, or week-over-week load changes — those signals are owned by weekly-coaching-patterns and activity-deep-dive. This skill emits coverage and imbalance gaps only.

Cap at 2 ideas. Emit-zero if none pass. Record discarded candidates (one line each) for Gate 4.

Gate 2 — Create proposals

For each idea, invoke /claude-code-hermit:proposal-create once:

Title: [<prefix>] <short idea title>
Evidence Source: capability-brainstorm
Evidence: <one paragraph: gap sentence + named grounding items>

Evidence Source: capability-brainstorm is intentional, not a copy-paste: it is the recurrence-bypass token recognized by proposal-create's three-condition rule (a brainstorm pass establishes the candidate, so condition 1 is waived). There is no separate domain-brainstorm token. Real provenance is carried by the tags below and the brainstorm-emit metrics event.

Set frontmatter: source: auto-detected, category: improvement, tags: [domain-brainstorm, ideation].

Parse the verdict:

• CREATE — note PROP-NNN.
• SUPPRESS — <code> — record suppression code; don't retry.
• DUPLICATE:<PROP-ID> — record existing ID; don't create.

After each verdict, append a metrics event (Node stdlib, no deps):

node -e "const fs=require('fs'); fs.appendFileSync('.claude-code-hermit/state/proposal-metrics.jsonl', JSON.stringify({ts:'<now ISO>',type:'brainstorm-emit',skill:'domain-brainstorm',verdict:'<CREATE|SUPPRESS|DUPLICATE>',proposal_id:'<PROP-NNN or null>'})+'\n','utf-8');"

Use the config.json timezone for <now ISO> (matching proposal-create), so this file isn't a mix of UTC and local timestamps. This event is what the kill-criteria audit reads — proposal-create's own created event does not carry per-skill provenance.

Do NOT invoke proposal-triage directly — /claude-code-hermit:proposal-create handles it.

Gate 3 — Emit batch message

Send one message per the Operator Notification protocol in CLAUDE.md.

Zero-emit:

🏋️ Domain brainstorm — 0 ideas emitted (<reason: thin context | all suppressed | all duplicates>)

Non-zero:

🏋️ Domain brainstorm (<N> idea(s))

1. **[prefix] <title>** — <one-line description>
   _Grounding: <item 1>, <item 2>_
   _Gap: <one-sentence friction>_
   PROP-NNN created  ·  (or: suppressed — <code>  ·  or: duplicate of PROP-NNN)

Gate 4 — Compiled artifact (non-empty runs only)

If ≥1 PROP was created (not suppressed or duplicate), write:

.claude-code-hermit/compiled/domain-brainstorm-YYYY-MM-DD-HHMM.md

---
title: Domain brainstorm — <ISO timestamp>
type: domain-brainstorm
created: <ISO timestamp with timezone>
tags: [domain-brainstorm, ideation]
source: interactive
proposals_created: [PROP-NNN, ...]
---

Body (150-line cap): ideas that passed (one paragraph each), discarded candidates (one line each), triage verdicts, inputs scanned (paths only, no content).

Do not tag foundational — this is a time-bounded ideation snapshot.

Zero-emit runs: skip the artifact entirely. Log one line to SHELL.md Findings: domain-brainstorm: 0 ideas emitted (<reason>)