AgentSkillsCN

based-intern-ops

运行 Based Intern 代理（执行测试、检查状态、本地运行）

SKILL.md
--- frontmatter
name: based-intern-ops
description: Operate the Based Intern agent (run tests, inspect state, run locally)
metadata: {"openclaw":{"os":["win32","linux","darwin"],"requires":{"anyBins":["node","npm","pnpm"]}}}

This skill helps you operate this repository's Based Intern agent.

Location:

  • •Repo root is {baseDir}/../.. (two levels up from this skill folder).

Safe defaults:

  • •Prefer read-only actions first (status, logs, state inspection).
  • •Never enable live trading unless explicitly requested and all guardrails are understood.
  • •Never print secret values (e.g. PRIVATE_KEY, API tokens). If you need to confirm env presence, list variable NAMES only.

Common actions (run from repo root):

  1. •Check health
  • •node -v
  • •npm -v
  • •npm test
  • •npm run lint
  • •npm run typecheck
  1. •Run the agent locally (safe)
  • •npm run dev
    • •Uses .env defaults; DRY_RUN=true, TRADING_ENABLED=false, SOCIAL_MODE=none unless you override.
  1. •Inspect persisted state
  • •Read data/state.json to see last post/trade times and circuit breakers.
  1. •Build + run production bundle
  • •npm run build
  • •npm start

Railway + OpenClaw Gateway workflow (remote execution)

Context:

  • •If you're connected to an OpenClaw Gateway running in Railway using Dockerfile.openclaw, the repo is at /app and the TypeScript project is at /app/based-intern.

Goals:

  • •Reproduce Railway-only behavior (Linux env, Railway-provided env vars)
  • •Run safe repo workflows: test/typecheck/build
  • •Inspect persisted state files if present

Important constraints:

  • •Don’t assume you can "edit code" in a running Railway container and keep it. Containers are immutable per deploy.
  • •Use this for execution/verification; make code changes via PRs/commits, then redeploy.

Railway-safe commands (do not leak secrets)

  1. •Identify deploy + environment
  • •cd /app/based-intern
  • •node -v && npm -v
  • •pwd && ls -la
  • •printenv | grep -E '^RAILWAY_' | sort || true
  • •printenv | cut -d= -f1 | sort | sed -n '1,200p'
  1. •Run repo workflows in Railway environment
  • •cd /app/based-intern && npm test
  • •cd /app/based-intern && npm run typecheck
  • •cd /app/based-intern && npm run build
  1. •Validate config without printing secrets
  • •cd /app/based-intern && node -e "try{require('./dist/src/config.js').loadConfig(); console.log('CONFIG_OK')}catch(e){console.error(String(e && e.message ? e.message : e)); process.exit(1)}"
  1. •Inspect runtime state (if available)
  • •cd /app/based-intern && ls -la data || true
  • •cd /app/based-intern && (test -f data/state.json && cat data/state.json | jq '.' || echo 'no data/state.json')
  1. •Agent logs
  • •For the actual Based Intern agent running as a Railway Worker, use Railway's service logs UI.

Attach to the running Railway agent (control server)

Prereqs:

  • •On the Railway agent service, set: CONTROL_ENABLED=true, CONTROL_PORT=8080, CONTROL_TOKEN=<strong secret>.
  • •From the Railway OpenClaw Gateway service, call the agent via Railway private networking.

If the agent Railway service name is basedintern, the internal base URL is typically:

  • •http://basedintern.railway.internal:8080

Read-only checks:

  • •curl -sS http://basedintern.railway.internal:8080/healthz
  • •curl -sS -H "Authorization: Bearer $CONTROL_TOKEN" http://basedintern.railway.internal:8080/status | jq .

Trigger a manual tick (safe; blocks if a tick is already running):

  • •curl -sS -X POST -H "Authorization: Bearer $CONTROL_TOKEN" "http://basedintern.railway.internal:8080/tick?reason=openclaw" | jq .

If you need to tailor runtime config, edit .env or set environment variables before launching the process.