Skip to content

Command-line reference

The installed entry point is sindri. Use sindri --help and the selected subcommand's --help to inspect the interface supplied by your runtime. Paths and IDs below are arguments you supply, not installed example locations.

Command Purpose Options
sindri validate-project STUDY_PATH Resolve and validate the three-document project without solving --help
sindri run STUDY_PATH Execute one new study-v2 project run --output-root PATH, --timeout SECONDS
sindri status RUN_DIR Read the last committed verified status --json
sindri cancel RUN_DIR Request cooperative local cancellation --reason TEXT
sindri resume RUN_DIR Recover an eligible interrupted format-2 run --timeout SECONDS
sindri replay RUN_DIR Re-execute a compatible terminal snapshot as a fresh run --output-root PATH, --timeout SECONDS
sindri evaluate score Assess and compare completed immutable candidates Repeat --candidate ID=RUN_DIR; --output PATH; optional approved --policy PATH, authorized --overrides PATH
sindri evaluate render Verify and re-render an evaluation in a new directory --assessment PATH, --output PATH
sindri evaluate prepare Prepare neutral evidence for a blind advisory review Repeat --candidate ID=RUN_DIR; --output PATH
sindri evaluate validate-review REVIEW_ROOT Validate immutable review history Optional --freeze-blind JSON, --agent-identity TEXT
sindri evaluate reconcile Copy frozen findings and disclose the matching assessment --blind-review PATH, --assessment PATH, --output PATH; optional --findings JSON
sindri study propose STUDY_PATH Prepare populated defaults without dispatch --nominal, --preset, --strategy, --setup, --spec, --overrides
sindri study start PROPOSAL_ID Confirm the exact plan and allowance Required --digest, --revision, --key; optional --queue-only, --parent-study
sindri study status STUDY_ID Read the plan, state and remaining allowance --help
sindri study cases STUDY_ID Read a bounded case page --offset, --limit
sindri study pause/resume/cancel STUDY_ID Control the exact current study Required --digest, --revision, --key
sindri study adcs REQUEST_JSON Prepare native ADCS sizing response from a verified nominal case Optional seed and prescribed experiment in the request
sindri study export STUDY_ID Capture a bounded verified local archive Required new --output and --key
sindri study verify-export ARCHIVE Verify an archive without extracting it Optional --manifest-sha256
sindri study service status/enable/disable Read or change local study availability Global hard disable remains authoritative
sindri export-schema --out-dir PATH Write the current JSON schema suite Required --out-dir
sindri adcs-42 validate RUN_DIR Validate indexed specialist handoff artifacts --help
sindri adcs-42 run RUN_DIR --case CASE_ID --executable PATH Execute one case in a separate workspace --workspace-root PATH, --timeout SECONDS (default 3600)
sindri cost refresh-fx --output PATH --currency CODE Prepare a validated ECB reporting-FX snapshot Repeat --currency; --timeout SECONDS (default 15); --force for replacement

Timeout values must be positive. A run, resume or replay returns nonzero when the resulting lifecycle is failed, incomplete or cancelled. Successful project validation or a completed run does not imply passing engineering assessments. status --json emits a structured bundle_unreadable error and exit status 2 when bundle integrity/loading fails.

Evaluation commands return 0 for a valid assessment even when engineering checks fail; invalid inputs return 2. They do not run solvers or call a model. Follow spacecraft evaluation for score interpretation, comparison compatibility and the independent two-pass review sequence.

Read run control before cancellation/recovery and specialist exchange before external execution or network FX preparation. There is no generic sweep or arbitrary external-import subcommand. Multi-case experiments use the shared study commands; see bounded design studies for exact confirmation, continuation, evidence and migration steps. Put --store PATH immediately after study to use a nondefault control store. Current authored study.yaml requires version 2.0.0; historical evidence keeps its original versions and bytes.