jaipilot-fast-executionlisted

# Run substantial Java commands efficiently Minimize end-to-end wall time, not merely one command's displayed duration. Preserve the controlling task's exact behavior, coverage, measurement, and final-verification requirements. Never skip work, weaken a gate, raise a timeout, change test semantics, or edit repository configuration just to make a run appear faster. ## Size work to the real machine 1. Read repository instructions, wrappers, build configuration, modules, CI commands, required services, configured test forks, and effective Java version. 2. Determine usable CPU and memory from the execution boundary. On Linux, inspect `getconf _NPROCESSORS_ONLN` and cgroup v2 `cpu.max` and `memory.max` when present; a container's quota matters more than host totals. 3. Choose one bounded worker count `N`: at least one, no greater than usable CPUs, and lower when build JVMs, test forks, services, or memory make saturation unsafe. Record `N` and non-secret resource evidence. 4. Classify substantial work as CPU-, memory-, I/O-, network-, or shared-state-bound. More workers are not faster when they create contention, throttling, or nondeterminism. ## Batch without corrupting evidence 1. Prefer one targeted `rg` traversal, one Git query with all pathspecs, batched reads, and one script invocation over repeated per-file subprocesses. 2. Run independent read-only discovery or isolated scripts concurrently up to `N`. Keep output attributable and propagate ever