test-plugin-harnessinglisted

Measure and optimise one phase of an oddyssey run - preflight, drive, observation, or the whole run - under one of the coding-agent CLIs the package is benchmarked on (opencode, claude, copilot), against main measured just before the work. Use when a phase is too slow or too expensive, when a harnessing change must be proven rather than asserted, when a run is suspected of composing work the package should ship, and before the PR of any change on the path of /odd-observe, /odd-verify or /odd-status. Drives the CLI headless, measures the phase, names where the time went, and separates what the package controls from provider latency. Never a substitute for launch-llms-benchmark, which grades findings; this grades the harness.
using-system/oddyssey · ★ 9 · Testing & QA · score 76

Install: claude install-skill using-system/oddyssey

# Testing a phase of the plugin's harness `launch-llms-benchmark` asks *how good is this model's report*. This asks a different question: **how much work does the package still make the model compose before it can do anything** — and it answers it for one named phase at a time, under one named CLI. Do not run the benchmark protocol for this; it grades findings, costs a full run, and its row is not what moves when a skill stops making the model write a script. The rules a change here must satisfy are `AGENTS.md`'s **Plugin harnessing** section. This skill is how you prove one landed — and that section says when a measurement is owed: before the PR of any change on the path of `/odd-observe`, `/odd-verify` or `/odd-status`. ## What you need before starting - **The phase**, named by the caller: `preflight`, `drive`, `observation`, or `whole`. Ask if it is not named — measuring the wrong phase wastes the run. - **The CLI** the runs are driven by: `opencode`, `claude` or `copilot` — the three `launch-llms-benchmark` runs, and the source of everything CLI-specific here: how a run is launched headless, the form of the model id it takes, where its session lives and how it is read. The scripts below implement those per CLI; the command's steps 3, 6 and 7 are the reference when one of them looks wrong. Default `opencode` — the fastest instrument on this table so far; measure on the CLI whose behaviour the change is about. - **The mission**, the same on every sampl