← CLI Reference

human agent

Agent development, migration, and compliance CLI

Develop, migrate, and validate HUMΛN agents from the terminal.

Migration workflow

# 1. Scan for existing agents
human agent detect .

# 2. Scaffold a HUMΛN wrapper (preserves your source file)
human agent import --from=langchain src/my-agent.py --output ./migrated

# 3. CI compliance gate
human agent check ./migrated/src/my-agent-agent.ts --strict

Install IDE migration skills: human setup cursor --agent-kit.

Framework guides: Migration hub.

---

human agent detect

Scan a file or directory for agent sources and frameworks.

OptionDescription
[path]File or directory (default: .)
--jsonJSON array for CI
--summaryOne-line counts only
**Output columns:** file · framework · compliance score (0–9) · recommended action

---

human agent import

Two modes:

  • Agent migration--from= scaffolds a HUMΛN handler, install-manifest.ts, prompts/system.md, tests, and README-migration.md.
  • Workflow import (default) — imports n8n/LangChain/etc. into HUMΛN Builder (WorkflowManifestV1).
  • Agent migration options

    OptionDescription
    --from openai, langchain, crewai, ecc, n8n, agentfield, custom, auto
    --output Output directory (default: ./migrated)
    --preserve-promptsCopy inline system prompts into prompts/system.md

    Workflow import options

    OptionDescription
    --localWrite manifest locally (no API)
    --platform Force platform detection
    --platformsList supported workflow platforms
    ---

    human agent check

    Run the 9-point HUMΛN Agent Compliance Checklist (static analysis).

    OptionDescription
    --strictTreat PARTIAL as FAIL
    --json{ score, overall, checks: [...] }
    --fixSuggest TODO markers (dry-run)
    --applyWrite --fix suggestions
    **Exit codes:** `0` pass · `1` fail · `2` config error (file missing)

    Checklist:

  • InstallManifest (@human/core)
  • ctx.prompts.load()
  • Prompt Defense Baseline
  • ctx.llm.complete({ prompt })
  • Delegation scope check (ctx.delegation.scope)
  • withProvenanceContext + ctx.provenance.log()
  • ctx.memory
  • No bare fetch/axios (prefer ctx.call)
  • handler({ execute }) from @human/agent-sdk
  • ---

    human agent docs

    Print documentation URLs for migration topics (migration, openai, langchain, sdk, …).

    human agent ctx

    Print the ExecutionContext (ctx.*) surface reference for handlers.

    ---

    Other agent commands

    CommandDescription
    human agent initScaffold a new agent project
    human agent devLocal dev server
    human agent deployDeploy to HUMΛN Cloud
    human agent catalogBrowse marketplace agents

    Marketplace publish checklist

    Before human marketplace publish:

  • Lint promptspnpm prompts:lint (Defense Baseline on system prompts)
  • Check #15human marketplace review --prompts-dir . (Prompt Safety Certification)
  • Full API reviewPOST /v1/marketplace/review with bundle path for checks 1–13 + 15
  • Listings expose prompt_certification (hashes + trust boundary) — not prompt bodies. See marketplace CLI docs.