Agents: command scripts with result references (CLI, MCP, app); parity estimate
- engine::script: parse scripts (lines, JSON lines, JSON arrays), resolve "$N.path" /
"${N.path}" references to earlier results, run with a failure report (+ tests)
- CLI: script [FILE|-] (headless or --connect to the running app; --save/--export,
--keep-going), app COMMAND [JSON] / --method, describe ID, commands FILTER, $N references in
run --cmd (+ integration tests)
- MCP batch: references and a script-text alternative (+ test)
- docs/agents.md; docs/parity.md scored over 275 catalogue features with cargo xtask parity;
ROADMAP: parity ~63% breadth / ~55% overall, ~600 agent-hours remaining
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
7145107896