honelisted

# Hone — Performance (Honing sharpens the edge — measured, incremental, never blind.) Read `${CLAUDE_PLUGIN_ROOT}/references/creed.md`, `${CLAUDE_PLUGIN_ROOT}/references/stacks.md`, and `${CLAUDE_PLUGIN_ROOT}/references/dispatch.md` first. Log: `bash ${CLAUDE_PLUGIN_ROOT}/scripts/ledger.sh append temper <slug> hone STARTED -` ## Process 1. **Baseline first.** If a prior `reports/test-perf.md` exists, this run diffs against it. If not, this run IS the baseline — say so; a first run produces numbers, not verdicts (nothing to regress against). 2. **Determine what to measure.** Pipeline mode: the operations the job's spec/plan touched. Standalone: ask ("which operation feels slow?"). Every measurement target is a concrete invocation, not a vibe. 3. **Detect tooling** (stacks.md): pick this skill's per-stack playbook (`references/{ts,python,go,java,rust}.md`) — TS/JS `node --cpu-prof` / vitest bench; Python cProfile / pytest-benchmark; Go `go test -bench` + pprof; Java JMH or JFR; Rust criterion / perf (always `--release`). Fallback: `time` on repeated invocations. 4. **Write the test brief** (`briefs/hone.md`): targets, tool invocations, **≥3 runs per measurement, report the MEDIAN** (single runs are noise), fixed inputs (document them so the next run compares like-for-like), report path `reports/test-perf.md`. 5. **Dispatch `smithy:temperer`** (routing role `testing`). Every number in the report must appear in tool output; profiles sav