plugins/claude-code-dev-hermit/skills/domain-brainstorm/SKILL.md
On-demand dev-voice brainstorm — reads codebase friction signals (git churn, test signal, manifest drift, README coverage) and emits at most 2 improvement ideas, each gated by proposal-triage before becoming a PROP. Invoke when the operator asks "what should I be fixing?", "anything wrong with X?", or "brainstorm improvements". Never runs autonomously.
npx skillsauth add gtapps/claude-code-hermit domain-brainstormInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
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.
Read these in parallel — all are cheap working-tree reads. Do not run the test suite or install tools.
Git activity
git log --oneline -50
git log --format= --name-only -50 | sort | uniq -c | sort -rn | head -20
Note the 5 most-churned files and any subsystem clusters.
Test signal
Read .claude-code-hermit/state/last-test.json (written by /dev-test). Extract status, exit_code, and duration_ms. If absent or last-modified >24h ago, note "no recent test run" — do not run the suite.
Manifest drift
ls for package.json, Cargo.toml, pyproject.toml, go.mod. For each found, check its lockfile (package-lock.json/yarn.lock/pnpm-lock.yaml, Cargo.lock, poetry.lock, go.sum) exists alongside it. Flag any mismatch.
README coverage
Read README.md and a one-level ls of the repo root. Note features or modules claimed in the README with no corresponding directory or file visible.
Think across all four inputs. For each candidate, both constraints must pass before including it:
git:cli/index.js, state:last-test.json duration=38s, file:package-lock.json missing, readme:§Feature X). These support the friction; friction is the bar.Map each passing idea to the closest dev prefix: [missing-tests], [tech-debt], [dependency], [tooling], or [architecture].
Cap at 2 ideas. Emit-zero if none pass. Record discarded candidates (one line each) for Gate 4.
For each idea, invoke /claude-code-hermit:proposal-create once:
Title: [<prefix>] <short idea title>
Evidence Source: capability-brainstorm
Evidence: <one paragraph: friction sentence + named grounding items>
Set frontmatter: source: auto-detected, category: improvement.
Evidence Source: capability-brainstormis reused intentionally: it is the only sourceproposal-triagerecognizes for the single-pass recurrence bypass, and adding a dedicateddomain-brainstormsource is a core edit this skill deliberately avoids. The cost: domain-brainstorm proposals carry capability-brainstorm provenance to any consumer that buckets by evidence source. Per-skill kill metrics are unaffected (they read thebrainstorm-emitevents below, which are taggedskill:"domain-brainstorm"). If this skill graduates from pilot, promote it to a first-class triage bypass source.
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 via the shared helper:
node "${CLAUDE_PLUGIN_ROOT}/scripts/append-metrics.js" .claude-code-hermit/state/proposal-metrics.jsonl '{"ts":"<now ISO>","type":"brainstorm-emit","skill":"domain-brainstorm","verdict":"<CREATE|SUPPRESS|DUPLICATE>","proposal_id":null}'
Set "proposal_id":"PROP-NNN" for CREATE events; use JSON null for SUPPRESS/DUPLICATE events.
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 — /proposal-create handles it.
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>_
_Friction: <one-sentence pain>_
PROP-NNN created · (or: suppressed — <code> · or: duplicate of PROP-NNN)
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>)
tools
Presence history & tracker-health report — current home/away state, reliability, recent arrival/departure transitions, and activity patterns for person/device_tracker entities. Use when the operator asks about presence history or when a presence-dependent automation (locks, alarm, vacuum, climate) misbehaves.
development
Evening house brief — end-of-day security check, device status, and energy snapshot. Runs as a daily routine at 22:30 or on demand.
tools
Browse and explain the hermit's Home Assistant automations — list by topic, filter by keyword with plain-language YAML explanations, or sort by last-fired. Read-only. Use when the operator asks "what automations do I have / what does this one do / which haven't fired."
tools
On-demand HA-voice brainstorm — reads entity inventory, automation/script listings, and operator intent to surface at most 2 capability-gap ideas, each gated by proposal-triage before becoming a PROP. Invoke when the operator asks "what automations am I missing?", "any coverage gaps?", or "brainstorm improvements". Never runs autonomously.