Runtime-behavior probes for Overwatch Workshop rules, and the captures recorded from the real client.
A probe is a small hand-written Workshop program. A human pastes it into the client, the whole Inspector text is saved verbatim, and a JSON record states the hypothesis, the observed values and a result of match, mismatch or inconclusive. The captures are the test fixtures for the scheduler simulator.
Each probe's measured result and run count are in probes/<id>/capture.json.
The scheduler simulator (crates/athena-sim) is a released library that depends on workshop-rs and builds for WebAssembly. It runs Workshop text, optionally with a JSON scenario, and returns a timeline trace. Wright consumes it by git tag. Its fixtures are the probe captures (cargo test). Placement is in docs/placement.md.
Use the workspace skill wrightkit-run-workshop-probe (.agents/skills/wrightkit-run-workshop-probe/SKILL.md). The agent prepares the probe and judges the result; you paste into the client and copy the Inspector text back. The steps and expected values (Chinese) are in docs/procedure.zh-CN.md.
To record a run from the Workshop Inspector log, run athena ingest probes/<id> <log-file>, or point athena watch <log-dir> at a directory of Inspector logs to ingest each finished run as it lands. From a checkout, cargo run -p athena-capture -- runs the athena binary (capture convention).
See docs/README.md: goal, architecture, simulator interface, scenario format, trace format, capture convention, findings, releases.