Output formats

--format selects the output of a report command: tty (default), json, sarif or markdown. The terminal and Markdown views are summaries for people; JSON is the exhaustive contract for tools.

Terminal

The combined cargo judge view groups findings by rule and shows one representative location per group, separated into evidence-backed findings and advisory heuristics, followed by next steps. --details lists every location. Colour follows --color auto|always|never; auto uses colour only on a terminal and honours NO_COLOR.

JSON

cargo judge --format json                       # .judge/judge.json
cargo judge dupes --format json --output reports/dupes.json

Every report command writes versioned JSON, by default to .judge/<command>.json, or to --output PATH. Every artifact starts with a header, followed by the command payload. The header records the working directory, the output path, a generation timestamp, a short command description and an assessment: informational, review_recommended, analysis_incomplete or blocking_findings. The assessment is a triage hint, not a refactoring instruction.

The combined report’s payload carries schema_version, analysis_universe (judge version, tier, targets, features, whether tests or generated code were included), counts, errors and the complete findings. The map, impact and progress contracts are covered by CLI tests, so tooling can rely on their schema versions.

SARIF

cargo judge --format sarif > judge.sarif

SARIF 2.1.0 is available on the report-producing commands and goes to stdout, ready for a code scanning upload.

Markdown

cargo judge --format markdown                          # .judge/judge.md
cargo judge compare .judge/baseline.json --format markdown

For the combined run, Markdown renders the grouped summary for a pull request or issue: an executive summary, evidence-backed and advisory groups with counts and representative locations, and next steps — never a raw per-finding dump. For compare it renders the delta table. See the Markdown example.

Progress events

cargo judge --progress judge-progress.jsonl

For the bare combined run, --progress PATH writes a flushed JSON Lines stream. Each record has schema_version, sequence, event (phase_started or phase_completed) and phase. It reports the analysis lifecycle only, never provisional findings, so an agent can follow a long run.