Skip to content

cli-reference.md

Latest commit

 

History

History
137 lines (103 loc) · 6.65 KB

cli-reference.md

File metadata and controls

137 lines (103 loc) · 6.65 KB

CLI Reference

Complete reference for all RuleProbe CLI commands and options.

ruleprobe parse <instruction-file>

Extract rules from an instruction file.

ruleprobe parse CLAUDE.md --format json
ruleprobe parse AGENTS.md --show-unparseable
ruleprobe parse AGENTS.md --llm-extract --show-unparseable
Option Default Description
--format <format> text Output format: json or text
--show-unparseable false Include lines that could not be converted to rules
--llm-extract false Send unparseable lines to an OpenAI-compatible API for additional extraction (requires OPENAI_API_KEY)

ruleprobe verify <instruction-file> <output-dir>

Parse rules from an instruction file. Verify agent output against them.

ruleprobe verify CLAUDE.md ./output --format text
ruleprobe verify AGENTS.md ./output --format summary --threshold 0.9
ruleprobe verify AGENTS.md ./output --agent claude --model opus-4 --format json --output report.json
ruleprobe verify AGENTS.md ./output --format detailed --severity error
ruleprobe verify AGENTS.md ./output --format ci
ruleprobe verify AGENTS.md ./output --format rdjson
ruleprobe verify AGENTS.md ./output --config ruleprobe.config.ts
ruleprobe verify AGENTS.md ./output --llm-extract
ruleprobe verify AGENTS.md ./output --rubric-decompose
ruleprobe verify AGENTS.md ./output --project tsconfig.json
ruleprobe verify AGENTS.md ./output --changed-since origin/main
Option Default Description
--format <format> text text, json, markdown, rdjson, summary, detailed, or ci
--threshold <number> 0.8 Compliance threshold (0-1) for pass/fail determination
--agent <name> unknown Agent identifier for report metadata
--model <name> unknown Model identifier for report metadata
--task <template-id> manual Which task template was used
--severity <level> all Filter: error, warning, or all
--output <path> stdout Write report to file
--config <path> auto-discovered Path to config file
--llm-extract false Run LLM extraction on unparseable lines (requires OPENAI_API_KEY)
--rubric-decompose false Decompose subjective rules via LLM (requires OPENAI_API_KEY)
--project <tsconfig> none tsconfig.json path for type-aware checks
--allow-symlinks false Follow symlinks outside the working directory
--changed-since <ref> none Only verify files changed since the given git ref (e.g., origin/main)

Format highlights: summary outputs a compact per-category table. detailed shows per-rule compliance percentages with evidence. ci produces key=value output with GitHub Actions ::error annotations. rdjson produces reviewdog-compatible JSON.

--changed-since <git-ref>: runs git diff --name-only --diff-filter=ACMR <ref>...HEAD (added, copied, modified, renamed; deletions excluded) and limits per-file checks to that set. Project-level rules (changelog-exists, strict-mode, etc.) still run. Cross-file rules like test-files-exist see the full file tree, so a missing test for a changed source file is still reported even when the test file itself is unchanged. The flag is opt-in: omit it and verify checks every file under <output-dir>. Requires git on PATH and the working directory to be inside a git repo with at least one commit. Exits with code 2 and a descriptive message if either is missing.

Exit codes: 0 all rules passed, 1 violations found, 2 execution error.

ruleprobe analyze <project-dir>

Discover all instruction files in a project, parse each, and report cross-file conflicts and redundancies.

ruleprobe analyze ./my-project --format text
ruleprobe analyze ./my-project --format json --output analysis.json
ruleprobe analyze ./my-project --semantic
ruleprobe analyze ./my-project --semantic --cost-report --semantic-log

Checks for CLAUDE.md, AGENTS.md, .cursorrules, .github/copilot-instructions.md, GEMINI.md, .windsurfrules, and .rules at the project root. Reports per-file rule counts, cross-file conflicts (same topic, contradictory instructions), redundancies (same rule in multiple files), and a category coverage map.

Option Default Description
--format <format> text Output format: text or json
--output <path> stdout Write report to file
--threshold <number> 0.8 Compliance threshold (0-1) for CI pass/fail
--semantic false Enable semantic analysis (requires ANTHROPIC_API_KEY)
--anthropic-key <key> none Anthropic API key (also: env ANTHROPIC_API_KEY or .ruleprobe/config.json)
--max-llm-calls <n> 20 Maximum LLM calls per semantic analysis
--no-cache false Disable profile caching
--semantic-log false Print semantic analysis log to stdout after results
--cost-report false Show token cost breakdown for semantic analysis

ruleprobe lint-config <instruction-file>

Parse an instruction file and emit an ESLint config. Flat config is the default; use --format legacy for .eslintrc.json output.

ruleprobe lint-config CLAUDE.md
ruleprobe lint-config CLAUDE.md --format legacy --output .eslintrc.json
ruleprobe lint-config AGENTS.md --format flat --output eslint.config.js
Option Default Description
--format <format> flat Output format: flat (ESLint flat config) or legacy (.eslintrc.json)
--output <path> stdout Write config to file

ruleprobe drift <md-file> <eslint-file>

Detect drift between an instruction file and an ESLint config. Reports rules present in only one side, severity mismatches, and argument differences.

ruleprobe drift CLAUDE.md .eslintrc.json
ruleprobe drift CLAUDE.md eslint.config.js --format markdown
ruleprobe drift AGENTS.md .eslintrc.json --format json --output drift-report.json
Option Default Description
--format <format> text Output format: text, json, or markdown
--output <path> stdout Write report to file

ruleprobe extract <eslint-file>

Parse an ESLint config and emit a markdown rules section suitable for pasting into an instruction file. Stylistic rules are reported but excluded from the output by default.

ruleprobe extract .eslintrc.json
ruleprobe extract eslint.config.js --output rules-section.md
Option Default Description
--output <path> stdout Write output to file