{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > how a job ends > ends `cancelled` when its signal aborted"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > how a job ends > ends `cancelled` when its signal aborted"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > how a job ends > ends `failed` as `INVALID_RESULT`, with its report, when the model submitted no result"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > how a job ends > ends `failed` as `INVALID_RESULT`, with its report, when the model submitted no result"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > how a job ends > ends `failed` as `LIMIT_REACHED` when the turn limit stopped the loop"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > how a job ends > ends `failed` as `LIMIT_REACHED` when the turn limit stopped the loop"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > how a job ends > ends `failed` as `PROVIDER_FAILURE` when the loop failed any other way"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > how a job ends > ends `failed` as `PROVIDER_FAILURE` when the loop failed any other way"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > how a job ends > removes a result file an earlier job left before the job starts"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > how a job ends > removes a result file an earlier job left before the job starts"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > how a job ends > tells the caller of each transcript event before passing it on, and gives the loop the job's signal"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > how a job ends > tells the caller of each transcript event before passing it on, and gives the loop the job's signal"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > passes the trusted job identity through the commands config into the harness"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > passes the trusted job identity through the commands config into the harness"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > the task's prompt-slot role > binds the task as `context` when the spec says so"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > the task's prompt-slot role > binds the task as `context` when the spec says so"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > the task's prompt-slot role > binds the task as `direct-command` when the spec says so"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > the task's prompt-slot role > binds the task as `direct-command` when the spec says so"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > the task's prompt-slot role > binds the task as `quote` when the spec says so"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > the task's prompt-slot role > binds the task as `quote` when the spec says so"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > the task's prompt-slot role > takes the role from the spec alone, whatever the task text says"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > the task's prompt-slot role > takes the role from the spec alone, whatever the task text says"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > with a fabric part > configures the spec's fabric session with its ceiling and passes its input cells"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > with a fabric part > configures the spec's fabric session with its ceiling and passes its input cells"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > with plain inputs and no fabric > gives the harness the spec's instructions and turn cap, and neither when the spec names none"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > with plain inputs and no fabric > gives the harness the spec's instructions and turn cap, and neither when the spec names none"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > with plain inputs and no fabric > leaves the model to the harness's own configuration when the spec names none"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > with plain inputs and no fabric > leaves the model to the harness's own configuration when the spec names none"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > with plain inputs and no fabric > leaves usage out of the report when the loop reports none"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > with plain inputs and no fabric > leaves usage out of the report when the loop reports none"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > with plain inputs and no fabric > passes no fabric session and no input cell"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > with plain inputs and no fabric > passes no fabric session and no input cell"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > with plain inputs and no fabric > reports usage including descendants when the loop totals it"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > with plain inputs and no fabric > reports usage including descendants when the loop totals it"}}
{"date":"2026-10-05","from":{"k":"unit","s":"cli","n":"runHarnessJob() > with plain inputs and no fabric > returns the submitted value, the job's handles, and its report"},"to":{"k":"unit","s":"agent-runner","n":"runHarnessJob() > with plain inputs and no fabric > returns the submitted value, the job's handles, and its report"}}
{"date":"2026-10-06","from":{"k":"unit","s":"agent-runner","n":"selectHarnessJobSandboxRuntime() > takes Docker by default off macOS"},"to":{"k":"unit","s":"agent-runner","n":"selectHarnessJobSandboxRuntime() > takes Docker by default off macOS and Linux"}}
{"date":"2026-10-07","from":{"k":"unit","s":"agent-runner","n":"selectHarnessJobSandboxRuntime() > takes Docker by default off macOS and Linux"},"to":{"k":"unit","s":"agent-runner","n":"selectHarnessJobSandboxRuntime() > refuses, as every job would be refused, off macOS and Linux, where no runtime is the default"}}