DataEcho — agent skills

Two Agent Skills for the DataEcho platform:

Skill	What it does
`dataecho`	Deploy anything to a live URL in seconds — a single file, a static site, or a full server-side app (any stack, via Dockerfile). Anonymous publish with a claim flow, incremental deploys, private Drives with scoped share tokens.
`dataecho-memory`	Persistent memory for agents — durable across sessions, machines, sandboxes, and platforms. A ~1 KB index + one fact per file on a private versioned drive; atomic concurrent-safe writes, full history/undo, one-command handoff to another agent.

Both are plain SKILL.md + self-contained scripts (bash + python3 stdlib, zero dependencies), so they work in any agent that supports the open Agent Skills format — Claude Code, OpenAI Codex, Gemini CLI, Cursor, GitHub Copilot / VS Code, Goose, opencode, Qwen Code, Amp, Cline, Kilo, Hermes, OpenClaw, and more.

Install

bash

# both skills, any supported agent (global; drop -g for project-local)
npx skills add mohocp/dataecho -g

# just one of them
npx skills add mohocp/dataecho --skill dataecho -g
npx skills add mohocp/dataecho --skill dataecho-memory -g

Claude Code plugin route:

Code

/plugin marketplace add mohocp/dataecho
/plugin install dataecho@dataecho

No npm? Install the deploy helper scripts directly:

Terminal

curl -fsSL https://dataecho.ai/install.sh | bash        # macOS / Linux

powershell

irm https://dataecho.ai/install.ps1 | iex               # Windows PowerShell

What's inside

skills/dataecho/SKILL.md — publish handshake, Dockerfile app contract, claim contract, Drives, API-key flow
- scripts/publish.sh — 3-call publish (create → upload → finalize), incremental deploys, anonymous-claim handling
- scripts/drive.sh — drive CRUD, import/export, scoped share tokens, ETag-safe writes
- scripts/publish.ps1 — Windows PowerShell publisher
skills/dataecho-memory/SKILL.md — the memory discipline (recall-first, update-don't-duplicate, forget-what's-wrong)
- scripts/memory.sh — recall / remember / forget / reindex / history / restore / handoff; every mutation commits fact + index in one atomic CAS batch

Security & privacy

Credentials: the API key (obtained via an email one-time code) is stored at ~/.artifact/credentials (chmod 600), or passed via $ARTIFACT_API_KEY. Nothing else is written outside the project. Revoke keys any time from the dashboard.
Network: scripts talk only to https://dataecho.ai (override: $ARTIFACT_BASE_URL). No telemetry, no third-party calls.
Anonymous publishes need no account; they expire in 24 h unless claimed. Claim tokens are saved locally under ~/.artifact/claims/.
Memory is private storage, scoped to your account; share access is via revocable, prefix/TTL-scoped drive tokens. Don't store secrets in memories — it's not a vault.
The curl | bash installer above is a convenience mirror of install.sh; the canonical scripts are versioned in this repo.

Links

Docs: https://dataecho.ai/docs · Agent context: https://dataecho.ai/llms.txt · llms-full.txt · OpenAPI: https://dataecho.ai/openapi.json · Discovery: https://dataecho.ai/.well-known/agent.json

License

MIT