AgentSkillsCN

chakravarti-cli

基于规范的智能体编排。创建规范、规划任务、运行作业、审核变更。

SKILL.md
--- frontmatter
name: chakravarti-cli
description: Spec-driven agent orchestration. Create specs, plan tasks, run jobs, and review changes.
license: MIT
compatibility: Claude Code, Cursor, any CLI-capable agent
metadata:
  version: "0.1.0"
  auto-generated: true
  generated-at: "2026-02-05T11:52:20Z"

Chakravarti CLI

Command-line interface for Chakravarti

Commands

ckrv cloud

Cloud execution commands.

Manage remote job execution via Chakravarti Cloud. Submit jobs, monitor progress, and retrieve results from cloud workers.

Subcommands: login, submit, status, cancel

bash
ckrv cloud

Examples:

Examples:

Login to cloud

ckrv cloud login

Submit a job

ckrv cloud submit my-feature

Check job status

ckrv cloud status <job-id>

ckrv cloud credentials

Manage git credentials for private repositories

bash
ckrv cloud credentials
ckrv cloud credentials add

Add a new git credential

bash
ckrv cloud credentials add [OPTIONS]

Options:

FlagDescription
--credential-typeCredential type (pat, deploy_key)
--nameName for this credential (e.g., "github-work")
--providerGit provider (github, gitlab, bitbucket, generic)
ckrv cloud credentials list

List stored credentials

bash
ckrv cloud credentials list
ckrv cloud credentials remove

Remove a credential

bash
ckrv cloud credentials remove <NAME>

Arguments:

NameRequiredDescription
nameYesName of the credential to remove

ckrv cloud login

Authenticate with Chakravarti Cloud

bash
ckrv cloud login [OPTIONS]

Options:

FlagDescription
--no-browserSkip opening browser automatically

ckrv cloud logout

Clear stored cloud credentials

bash
ckrv cloud logout [OPTIONS]

Options:

FlagDescription
--force, -fForce logout without confirmation

ckrv cloud whoami

Display current authenticated user

bash
ckrv cloud whoami

ckrv diff

View changes between current branch and base.

Shows a summary of modified, added, and deleted files compared to the base branch. Helps verify what will be included in a pull request.

Output can be formatted as JSON for programmatic use.

bash
ckrv diff [OPTIONS]

Options:

FlagDescription
--base, -bBase branch to compare against (default: main or master)
--colorColor mode for diff output
--filesShow file list only
--statShow diff statistics only
--summaryGenerate AI summary of changes

Examples:

Examples:

Show diff summary

ckrv diff

Show diff against specific branch

ckrv diff --base main

Output as JSON

ckrv diff --json

ckrv fix

Fix verification errors with AI.

Analyzes failed tests, lint errors, or build issues and uses AI to automatically generate fixes. Runs in an isolated Docker sandbox.

Best used after ckrv verify identifies issues.

bash
ckrv fix [OPTIONS]

Options:

FlagDescription
--checkRe-run verification after fixing
--errorSpecific error message to fix (from UI)
--lintFix only lint errors
--testFix only test failures
--typecheckFix only type errors

Examples:

Examples:

Fix all errors

ckrv fix

Fix with specific agent

ckrv fix --agent claude-3.5

Fix only test failures

ckrv fix --tests-only

ckrv init

Initialize Chakravarti in the current repository.

Creates the .chakravarti/ directory with default configuration files including config.yaml for project settings and initializes the specs directory.

This is typically the first command to run when setting up a new project for AI-driven development with Chakravarti.

bash
ckrv init [OPTIONS]

Options:

FlagDescription
--forceForce reinitialization even if already initialized

Examples:

Examples:

Initialize in current directory

ckrv init

Initialize with verbose output

ckrv init --verbose

ckrv logs

Stream or view logs from a cloud job.

Shows real-time output from running jobs or historical logs from completed jobs. Supports filtering by task or agent.

Use --follow for continuous streaming.

bash
ckrv logs <JOB_ID> [OPTIONS]

Arguments:

NameRequiredDescription
job_idYesJob ID to get logs for

Options:

FlagDescription
--follow, -fFollow log output (stream in real-time)
--tail, -nNumber of recent log lines to show (default: 100)

Examples:

Examples:

Stream logs from running job

ckrv logs <job-id> --follow

View completed job logs

ckrv logs <job-id>

Filter by task

ckrv logs <job-id> --task 3

ckrv plan

Generate execution plan from tasks using AI.

Analyzes the specification and tasks file to create a detailed implementation plan. Runs in a Docker container for isolation.

The plan breaks down work into atomic steps that AI agents can execute.

bash
ckrv plan [SPEC] [OPTIONS]

Arguments:

NameRequiredDescription
specNoPath to the specification directory. If not provided, will detect from branch name

Options:

FlagDescription
--force, -fForce regeneration even if plan.yaml already exists

Examples:

Examples:

Generate plan for a specification

ckrv plan my-feature

Generate plan with GLM model

ckrv plan my-feature --model glm-4.7

Skip confirmation prompt

ckrv plan my-feature --yes

ckrv promote

Create a pull request for the current branch.

Pushes the feature branch and creates a pull request on GitHub/GitLab. Auto-generates PR title and description from the specification.

Requires remote repository access and appropriate permissions.

bash
ckrv promote [OPTIONS]

Options:

FlagDescription
--base, -bTarget branch for the PR (default: main or master)
--draftCreate as draft PR
--openOpen PR URL in browser after creation
--pushPush branch to remote before creating PR
--remoteRemote name (default: origin)
--skip-verifySkip verification checks

Examples:

Examples:

Create PR with auto-generated description

ckrv promote

Create as draft PR

ckrv promote --draft

Create PR with custom title

ckrv promote --title "feat: add user auth"

ckrv pull

Pull results from a completed cloud job.

Downloads all changes made during cloud execution and applies them to the local repository. Creates or updates the feature branch.

Jobs must be in a 'completed' state to pull.

bash
ckrv pull <JOB_ID> [OPTIONS]

Arguments:

NameRequiredDescription
job_idYesJob ID to pull results from

Options:

FlagDescription
--applyApply diff to current worktree (default: true)
--outputOutput diff to file instead of applying

Examples:

Examples:

Pull results to current directory

ckrv pull <job-id>

Pull and create new branch

ckrv pull <job-id> --branch feature/new

ckrv qa

QA code review and bug analysis.

AI-powered code review and quality assurance. Analyzes changes for potential bugs, security issues, and code quality improvements.

Subcommands: review, bugs, report

bash
ckrv qa

Examples:

Examples:

Review current changes

ckrv qa review

Analyze for bugs

ckrv qa bugs

Generate QA report

ckrv qa report

ckrv qa bugs

Analyze for potential bugs

bash
ckrv qa bugs [OPTIONS]

Options:

FlagDescription
--baseBranch to compare against (default: main)

ckrv qa report

Generate full QA report

bash
ckrv qa report [OPTIONS]

Options:

FlagDescription
--baseBranch to compare against (default: main)
--fullInclude all analysis types
--output, -oOutput file path

ckrv qa review

Review code quality of changes

bash
ckrv qa review [OPTIONS]

Options:

FlagDescription
--baseBranch to compare against (default: main)
--output, -oOutput file path

ckrv run

Run a job based on a specification.

Executes the implementation plan using AI agents in isolated Docker sandboxes. Each task is executed in sequence with full logging and progress tracking.

Results are committed to a feature branch for review.

bash
ckrv run [SPEC] [OPTIONS]

Arguments:

NameRequiredDescription
specNoPath to the specification file. If not provided, will detect from branch name

Options:

FlagDescription
--agentAgent to use for execution: claude or codex
--cloudExecute job in Chakravarti Cloud instead of locally
--credentialGit credential name to use for cloud execution (for private repos)
--executor-model, -eOverride the AI model/agent to use for execution
--optimize, -oOptimization strategy

Examples:

Examples:

Run all tasks for a specification

ckrv run my-feature

Run with specific agent

ckrv run my-feature --agent claude-3.5

Dry run (show what would be done)

ckrv run my-feature --dry-run

ckrv spec

Create or manage feature specifications.

Specifications are the source of truth for AI-driven development. They define what needs to be built, the requirements, and acceptance criteria.

Subcommands: new, list, validate, edit, show

bash
ckrv spec

Examples:

Examples:

Create a new specification

ckrv spec new "Add user authentication"

List all specifications

ckrv spec list

Validate a specification

ckrv spec validate my-feature

ckrv spec clarify

Resolve clarifications in an existing spec

bash
ckrv spec clarify [SPEC]

Arguments:

NameRequiredDescription
specNoPath to the spec file (optional - auto-detects from current branch if not provided)

ckrv spec design

Generate technical design document from a specification

bash
ckrv spec design [SPEC] [OPTIONS]

Arguments:

NameRequiredDescription
specNoPath to the spec file (optional - auto-detects from current branch if not provided)

Options:

FlagDescription
--force, -fForce regeneration of design even if it exists

ckrv spec init

Initialize an empty spec directory with templates

bash
ckrv spec init <NAME>

Arguments:

NameRequiredDescription
nameYesName for the new spec directory

ckrv spec list

List all specifications

bash
ckrv spec list

ckrv spec new

Create a new specification using AI from a natural language description

bash
ckrv spec new <DESCRIPTION> [OPTIONS]

Arguments:

NameRequiredDescription
descriptionYesNatural language description of the feature (e.g., "Add user authentication")

Options:

FlagDescription
--name, -nOptional short name for the spec (auto-generated from description if not provided)

ckrv spec tasks

Generate implementation tasks from a specification

bash
ckrv spec tasks [SPEC] [OPTIONS]

Arguments:

NameRequiredDescription
specNoPath to the spec file (optional - auto-detects from current branch if not provided)

Options:

FlagDescription
--force, -fForce regeneration of tasks even if they exist

ckrv spec validate

Validate a specification file

bash
ckrv spec validate [PATH]

Arguments:

NameRequiredDescription
pathNoPath to the spec file (optional - auto-detects from current branch if not provided)

ckrv test

Run tests in sandbox, plan and write new tests.

Comprehensive test management with AI assistance. Can run existing tests, analyze coverage gaps, and generate new tests using AI agents.

Subcommands: run, plan, write

bash
ckrv test

Examples:

Examples:

Run all tests

ckrv test run

Plan tests for uncovered code

ckrv test plan

Write new tests with AI

ckrv test write --agent claude-3.5

ckrv test coverage

Check test coverage of changed files

bash
ckrv test coverage [OPTIONS]

Options:

FlagDescription
--baseBranch to compare against (default: main)

ckrv test plan

Analyze changes and generate test plan

bash
ckrv test plan [OPTIONS]

Options:

FlagDescription
--baseBranch to compare against (default: main)

ckrv test run

Run existing tests in sandbox

bash
ckrv test run [OPTIONS]

Options:

FlagDescription
--baseBranch to compare against (default: main)

ckrv test write

Write new tests using test writer agent

bash
ckrv test write [OPTIONS]

Options:

FlagDescription
--baseBranch to compare against (default: main)
--runRun tests after writing

ckrv ui

Start the Web UI dashboard.

Launches a local web server providing a visual interface for managing specifications, viewing execution progress, and reviewing AI agent output.

Opens automatically in your default browser.

bash
ckrv ui [OPTIONS]

Options:

FlagDescription
--portPort to listen on (default: 3000)

Examples:

Examples:

Start UI on default port

ckrv ui

Start on custom port

ckrv ui --port 8080

Don't open browser automatically

ckrv ui --no-open

ckrv verify

Run tests, lint, and quality checks.

Validates the current code against project quality standards. Runs the test suite, linters, and any custom verification scripts.

Failed verifications can be fixed with ckrv fix.

bash
ckrv verify [OPTIONS]

Options:

FlagDescription
--continue-on-failureContinue on failure (run all checks even if some fail)
--fixAuto-fix issues where possible
--lintRun only lint checks
--saveSave results to verification.yaml
--testRun only tests
--typecheckRun only type checks

Examples:

Examples:

Run all verifications

ckrv verify

Run only tests

ckrv verify --tests-only

Run in JSON output mode

ckrv verify --json

Global Options

These options apply to all commands:

FlagDescription
--jsonOutput format: JSON instead of human-readable
--quiet, -qSuppress non-essential output
--verbose, -vEnable verbose logging
--help, -hPrint help
--version, -VPrint version