@agentskb/cli
Advanced tools
CLI tool for AgentsKB Cognitive Control Plane - Generate project manifests and lockfiles for deterministic AI development
The Cognitive Control Plane for AI Agents
Generate project manifests and lockfiles for deterministic, reproducible AI development.
AgentsKB CLI transforms AI agent development from non-deterministic trial-and-error to reproducible, verified builds.
Like package.json defines dependencies and package-lock.json locks versions, AgentsKB uses:
agentskb.yaml - Defines project requirements and knowledge planagentskb.lock - Locks verification IDs for deterministic buildsAI agents are non-deterministic:
Engineering Managers cannot ship non-deterministic systems.
AgentsKB CLI provides:
npm install -g @agentskb/cli
Or use with npx:
npx @agentskb/cli init
cd my-project
agentskb init --stack "Next.js, Supabase, Stripe"
This generates agentskb.yaml with:
Example output:
🚀 AgentsKB Project Initialization
Decomposed into 47 questions
📊 Readiness Analysis:
✓ 89% Ready (Excellent!)
Coverage: 89.4%
42 verified / 47 total questions
📋 Implementation Phases:
✓ Foundation & Architecture (Ready)
✓ Core Features (Ready)
◐ Security & Compliance (Partial)
✓ Deployment & Monitoring (Ready)
✓ Manifest created: agentskb.yaml
agentskb check
Shows:
agentskb lock
Creates agentskb.lock with pinned verification IDs for reproducibility.
Commit both files to git:
git add agentskb.yaml agentskb.lock
git commit -m "Lock AgentsKB knowledge versions"
agentskb initInitialize a new AgentsKB project.
Options:
--stack <technologies> - Technology stack (comma-separated)--requirements <requirements> - Project requirements (comma-separated)--name <name> - Project name (defaults to directory name)--no-interactive - Run in non-interactive modeExamples:
# Interactive mode
agentskb init
# Non-interactive mode
agentskb init --stack "Next.js, PostgreSQL" --requirements "Auth, Payments"
agentskb checkCheck project readiness and coverage.
What it shows:
Example output:
📊 AgentsKB Readiness Check
Overall Readiness:
🎯 89% - Excellent! Ready to build
42/47 questions verified (89.4%)
Phase Breakdown:
✓ Foundation & Architecture
Status: Ready | 15/15 (100%)
◐ Security & Compliance
Status: Partial | 10/12 (83%)
⨯ What is SOC2 audit logging in PostgreSQL?
⨯ How to implement RBAC with Supabase RLS?
agentskb lockGenerate lockfile for deterministic builds.
What it does:
Benefits:
agentskb update(Coming soon) Update lockfile to latest knowledge versions.
After running agentskb lock, inject verified knowledge into your AI agent:
import yaml
from langchain.prompts import ChatPromptTemplate
from langchain.chat_models import ChatOpenAI
from langchain.schema import SystemMessage, HumanMessage
# Load lockfile
with open('agentskb.lock') as f:
lockfile = yaml.safe_load(f)
# Build system prompt from verified questions
verified_knowledge = "\n".join([
f"- {q['question']}"
for q in lockfile['questions']
])
system_prompt = f"""You are an expert developer with verified knowledge of:
{verified_knowledge}
IMPORTANT: Use ONLY this verified knowledge. If asked about topics outside
this scope, respond: "I don't have verified knowledge about that yet."
"""
# Use with ChatOpenAI
chat = ChatOpenAI(temperature=0, model="gpt-4")
messages = [
SystemMessage(content=system_prompt),
HumanMessage(content="How do I use useActionState in Next.js 15?")
]
response = chat(messages)
print(response.content)
import { generateText } from 'ai';
import { openai } from '@ai-sdk/openai';
import { readFileSync } from 'fs';
import yaml from 'js-yaml';
// Load lockfile
const lockfile = yaml.load(readFileSync('agentskb.lock', 'utf8'));
const questions = lockfile.questions.map(q => `- ${q.question}`).join('\n');
// Generate with verified knowledge
const { text } = await generateText({
model: openai('gpt-4'),
system: `You have verified knowledge of:\n${questions}\n\nUse ONLY verified knowledge.`,
prompt: userQuestion
});
console.log(text);
import OpenAI from 'openai';
import yaml from 'js-yaml';
import { readFileSync } from 'fs';
const openai = new OpenAI();
const lockfile = yaml.load(readFileSync('agentskb.lock', 'utf8'));
const completion = await openai.chat.completions.create({
model: "gpt-4",
messages: [
{
role: "system",
content: `Verified knowledge:\n${JSON.stringify(lockfile.questions, null, 2)}`
},
{ role: "user", content: userQuestion }
]
});
console.log(completion.choices[0].message.content);
import yaml
import anthropic
# Load lockfile
with open('agentskb.lock') as f:
lockfile = yaml.safe_load(f)
# Build knowledge context
knowledge = "\n".join([f"- {q['question']}" for q in lockfile['questions']])
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-3-5-sonnet-20241022",
max_tokens=1024,
system=f"You have verified knowledge of:\n{knowledge}\n\nUse ONLY this verified knowledge.",
messages=[
{"role": "user", "content": user_question}
]
)
print(message.content[0].text)
💡 Tip: For production use, fetch the actual answers from AgentsKB API using the verification_id for full context, not just the questions.
agentskb.yaml (Project Manifest)Defines project requirements and auto-generates knowledge plan.
project: "Ticket Management App"
version: "1.0"
stack:
- "Next.js 15 (App Router)"
- "Supabase (PostgreSQL + Realtime)"
- "Stripe (Payments)"
requirements:
- "Real-time ticket updates"
- "SOC2 compliance"
repositories:
- name: "default"
url: "https://api.agentskb.com"
scope: "public"
# AUTO-GENERATED BY AGENTSKB
plan:
total_questions: 47
coverage: 0.89
readiness_score: 89
phase_1_foundation:
status: "ready"
questions:
- id: "db-schema-001"
question: "What database schema for ticket management?"
verified: true
confidence: 0.990
constraints:
enforced:
- rule: "DO_NOT_USE_getInitialProps"
reason: "Incompatible with Next.js App Router"
severity: "blocking"
dependencies:
graph:
- from: "db-schema-001"
to: ["auth-jwt-001"]
type: "prerequisite"
agentskb.lock (Lockfile)Pins verification IDs for reproducibility.
# agentskb.lock (auto-generated - DO NOT EDIT)
lockfile_version: "1.0"
locked_at: "2025-11-21T14:30:00Z"
project: "Ticket Management App"
questions:
- id: "db-schema-001"
verification_id: "a3f9d8e2b1c4" # Hash of specific answer version
question: "What database schema for ticket management?"
confidence: 0.990
verified_at: "2025-11-15T09:12:00Z"
sources:
- "https://www.postgresql.org/docs/current/datatype.html"
- "https://supabase.com/docs/guides/database"
checksum: "7f3e8d9a..." # Integrity verification
AGENTSKB_API_KEY - Your AgentsKB API key (optional for free tier)OPENAI_API_KEY - OpenAI API key for task decomposition (optional, uses mock if not set)When you run agentskb init, the CLI:
The CLI queries AgentsKB's knowledge base (3,276+ Q&As at 0.990 confidence) to determine:
When you run agentskb lock, the CLI:
Result: Reproducible builds that work the same today as 6 months from now.
AgentsKB CLI generates files that AI agents (Claude Code, Cursor, GitHub Copilot) can read via MCP:
agentskb.yaml to understand project requirementsagentskb.lock for reproducibilityComing soon: Direct MCP integration for automatic manifest reading.
No! The CLI works with anonymous access (30 questions/month). For higher limits, sign up for free (100/month) or upgrade to Pro/Scale.
The CLI uses a mock decomposition for demo purposes. For production use, set OPENAI_API_KEY for GPT-4 powered task decomposition.
Update when:
Run agentskb update (coming soon) and review changes in git before committing.
Yes! Edit the top section (project, stack, requirements). The plan section is auto-generated - regenerate with agentskb init.
Never edit manually! It's auto-generated. Run agentskb lock to regenerate.
See examples/ directory for:
We welcome contributions! See CONTRIBUTING.md for guidelines.
MIT © AgentsKB Team
Built with ❤️ by the AgentsKB Team
Making AI development deterministic, one manifest at a time.
FAQs
AgentsKB CLI - Researched answers to any problem. Your AI, upgraded. Lock answers, check coverage, verify facts.
The npm package @agentskb/cli receives a total of 58 weekly downloads. As such, @agentskb/cli popularity was classified as not popular.
We found that @agentskb/cli demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Company News
Open source maintainers are under more pressure than ever. We're raising our open source program from the Team plan to the Business plan, free.
Security News
The supply chain control that delays freshly published gems now covers lockfile generation and gem vendoring in Ruby projects.
Security News
During a UK cyber test, a Mythos 5 agent used sockpuppets, social engineering, and prompt injection to try to get a maintainer to merge malware.