Commands

What it reports

• The stack, framework and file count it detected
• A menu covering review, check, watch, architecture, impact, map, tests, dead code and more
• Falls back to `little-owl check` when output is piped or running in CI

Useful options

-C, --cwd <dir>

Run against another directory.

What it reports

• What it detected: stack, framework, file count, git
• What it will watch for, in plain words, before it writes anything
• A config file, a baseline, and LITTLE_OWL.md — the briefing file AI assistants read
• One recommended next command

Useful options

-i, --interactive

Choose the structure and strictness yourself instead of having them detected.

--force

Overwrite an existing configuration.

--no-baseline

Skip creating a baseline.

--no-agent-file

Skip writing LITTLE_OWL.md.

What it reports

• A one-line health score, then a verdict: is this fine, or does it need work?
• Issue counts by priority — critical, important, minor — each with what that level means
• The first few issues, numbered, with the file, the line and a plain-language summary
• One recommended next command
• A warning when the scan was partial, so a truncated run never reads as a clean one

Useful options

--all

List every issue, not just the first few.

--details

The full explanation of every issue, with rule ids.

--json

Machine-readable output on a versioned schema.

-q, --quiet

Only the essentials.

--no-cache

Ignore the parse cache.

What it reports

• For an issue: what happened, why it matters, where it is, the related files, what should happen instead, how to fix it and how to confirm the fix
• Any term a non-expert might not know, defined in one clause
• For a file: the commit that introduced it, messages that record a reason, who maintains it and what imports it today
• An evidence rating of strong, partial or none — and a plain statement when the history is silent

Useful options

--technical

Include the rule id and the raw evidence.

--json

Machine-readable output.

What it reports

• The files involved: where the change goes, and what may need updating with it
• The goal, in one sentence
• Two ways to apply it — give the brief to your assistant, or do it yourself
• A markdown brief with the file, line, function, related files, risks and acceptance criteria

Useful options

--brief

Print only the brief, ready to pipe or copy.

--json

Machine-readable output.

What it reports

• Which issues are fixed, and which are still there
• Anything new that appeared since — a fix that trades one problem for another is not finished
• How the health score moved
• Exit code 1 while the issue you named is still reproducing, so a script can wait on it

Useful options

--tests

Also run this project's own test command.

--json

Machine-readable output.

What it reports

• The loop, and which command answers which question
• Your layers, the size limits this project agreed to, and how to read a priority
• What not to touch — in particular that editing the baseline to silence a finding is not a fix

Useful options

--force

Overwrite an existing file. Without it, an edited copy is left alone.

What it reports

• Changed files with lines added and removed
• Every score before and after, and the concrete counts behind each move
• Only the findings this change introduced — existing debt is not repeated at you
• Files changed outside the area you scoped the work to

Useful options

-b, --base <ref>

Git ref to compare against.

-s, --scope <glob>

Area the change was meant to touch. Repeatable.

--prompt

Print an AI prompt instead of the report.

--details

Show every finding.

--json

Machine-readable output.

What it reports

• The files you just changed, and how many modules import them
• New findings grouped by whether they are in the changed files, in files that depend on them, or elsewhere
• Score movement against a fixed reference, not against the state a second ago

Useful options

--debounce <ms>

Delay before re-analysing. Default 400.

--prompt

Include an AI prompt with each report.

What it reports

• Current health, with the previous baseline alongside it when one exists
• How many findings appeared since the last baseline, before you accept them
• The written baseline file, which belongs in version control

Useful options

-y, --yes

Write without asking.

--show

Print the existing baseline instead of writing one.

--json

Machine-readable output.

What it reports

• The last runs with their overall score and whether each improved or degraded
• Baseline snapshots, marked as such

Useful options

-n, --limit <count>

How many entries to show. Default 10.

--json

Machine-readable output.

What it reports

• Each issue with its priority, file, line, enclosing function and related files
• Current behaviour, expected behaviour, the risks, and acceptance criteria
• Constraints — fix only what is named, change no behaviour, add no dependencies, never edit the baseline to silence a finding
• The commands to run when the work is done

Useful options

--compact

A short numbered list instead of the full brief.

-s, --scope <glob>

Restrict the assistant to these paths.

--all

Include findings that predate this change.

-n, --max <count>

Maximum number of issues to include.

What it reports

• A one-line status, the finding counts and the score movement
• The findings that drove the verdict
• Exit code 0 or 1, and a PARTIAL ANALYSIS line if the scan did not cover everything

Useful options

-b, --base <ref>

Git ref to compare against.

--fail-on <level>

error | warning | never.

--max-drop <points>

Largest acceptable drop in the overall score.

--all

Consider pre-existing findings too, not just new ones.

--json

Machine-readable output.

What it reports

• A suggested reading order — where execution starts, then what most code depends on
• Areas with their size, and how many imports cross into each one
• Detected layers, entry points with their route paths, and the modules most files depend on
• External services grouped by what they are, rather than by package name

Useful options

--json

Machine-readable output.

What it reports

• Affected files ranked by how many import hops away they are
• Route-like entry points among them, labelled with their URL
• Tests that reach the change, and external packages it talks to
• A risk level, and a lowered confidence when a dynamic import could reach further than it can see

Useful options

-f, --files <paths>

Additional files to analyse.

-b, --base <ref>

Git ref to compare against.

--json

Machine-readable output.

What it reports

• Candidates graded high, medium or low confidence
• Why each one looks unused, and what undermines that conclusion
• Exported names nothing imports, for files that are otherwise in use
• Files skipped because a framework convention makes them entry points

Useful options

--min-confidence <level>

high | medium | low.

--include-tests

Consider test files too.

--no-cache

Analyse without writing anything to the project.

--json

Machine-readable output.

What it reports

• Modules with real logic that no test file reaches through imports
• Modules a test reaches without naming every exported behaviour
• Files skipped because they are configuration, scripts or migrations

Useful options

--changed

Only look at what the current change touched.

-b, --base <ref>

Git ref to compare against.

--json

Machine-readable output.

What it reports

• The layer chain, and whether it was configured or inferred
• Boundary violations and skipped layers
• Circular dependencies, with the whole loop named

Useful options

--json

Machine-readable output.

What it reports

• Packages imported but not declared
• Packages declared but never imported, with build tooling filtered out
• It is hygiene, not security — run your package manager's audit for vulnerabilities

Useful options

--json

Machine-readable output.

What it reports

• Which config file was loaded, or that defaults are in use
• Layers, thresholds and CI settings as resolved
• Every rule with its active severity, with --rules

Useful options

--rules

List every rule and its severity.

--json

Machine-readable output.

What it reports

• Node version, detected stack, git availability, config and baseline status
• How many files were analysed, and whether the scan was cut short
• Unresolved imports, detected layers and test files found

Useful options

--json

Machine-readable output.