dada

hosted backend infra for openclaw agents. managed databases, webhooks, and file hosting — so your agent can focus on the work, not the plumbing.

Security Scan

The skill's declared purpose (hosted backend infra via a CLI) matches the instructions: it tells the agent to invoke a published CLI (npx or GitHub release) to manage projects, tables, webhooks and files — nothing in the instructions requests unrelated secrets or system-wide privileges.

Purpose & Capability

The name/description align with the CLI-centric actions described in SKILL.md (create projects, tables, webhooks, file hosting). Minor mismatch: SKILL.md recommends using npx (requires Node.js) but the registry metadata lists no required binaries; also SKILL.md implies a local keypair is stored on disk but no config path was declared.

Instruction Scope

Instructions are narrowly scoped to running the dada CLI (project/table/record/webhook commands), streaming webhook SSE, and submitting feedback. They do not ask the agent to read arbitrary files, environment variables, or unrelated system configuration. Note: webhook watch is a long-running network stream and the login flow creates a local Ed25519 keypair (agent will read/write that local file).

Install Mechanism

No install spec in the registry (instruction-only), but SKILL.md directs use of 'npx @usedada/cli' or downloading a binary from GitHub Releases. Both are common for a CLI; npx executes remote npm code at runtime (inspect @usedada/cli source before running). GitHub Releases is an acceptable release host.

Credentials

The skill declares no required environment variables or primary credential, which is consistent with the instructions. However, the CLI establishes a persistent Ed25519 identity stored locally (file I/O) and requires network access to the hosted backend; those effects are not represented by declared config paths or binaries and should be confirmed before use.

Persistence & Privilege

The skill is not forced-always and does not request elevated/agent-wide persistence. It asks the agent to run an external CLI which will create a local keypair and keep network sessions (SSE) as needed; this is consistent with the tool's purpose.

Assessment

This skill appears coherent for a hosted backend CLI, but take these precautions before installing or running it: (1) confirm where the CLI stores the Ed25519 keypair (path and file protections) and whether you’re comfortable the agent will create/read that file; (2) ensure Node.js/npx is available if you plan to use 'npx' (or prefer downloading and vetting the GitHub release binary first); (3) review the @usedada/cli source or the release binary before executing to ensure it doesn't perform unexpected actions; (4) be aware webhook-watch opens long-lived network connections and webhook URLs are sensitive — store them securely; (5) if you need stricter guarantees, ask the publisher for explicit config-paths, storage locations, and a reproducible audit of the CLI behavior.

Use dada for persistent structured storage, webhooks, and file hosting. Each project gets its own isolated database with typed schemas. All operations go through the CLI.

Install

npx (requires Node.js):

npx @usedada/cli

Or download a prebuilt binary from GitHub Releases: https://github.com/honeybadge-labs/dada/releases

If you installed a binary directly, all commands below work with just dada instead of npx @usedada/cli.

Setup (once)

First login creates your identity (Ed25519 keypair stored locally):

dada login --nickname myagent --email [e-mail]

Subsequent logins just reconnect (keypair preserved on disk):

dada login

Common Commands

Projects

• dada project create <name>: create a new project
• dada project list: list all projects
• dada project use <name>: set active project

Tables

• dada table create <name> <field:type ...>: create a table with typed fields
• dada table list: list tables in active project
• dada table describe <name>: show table schema

Records

• dada insert <table> '<json>': insert a single record
• dada bulk-insert <table> '[<json>, ...]': insert multiple records in one request (preferred for batches)
• dada query <table> [-w filter]: query records with optional filters
• dada update <table> '<json>' -w filter: update matching records
• dada delete <table> -w filter: delete matching records

Webhooks

• dada webhook create <name>: create an inbound webhook (returns URL)
• dada webhook list: list all webhooks in active project
• dada webhook delete <name>: delete a webhook
• dada webhook watch <name>: stream webhook events via SSE (Ctrl+C to stop)
• dada webhook dequeue <name> [--limit N]: dequeue queued events (default limit: 100)

Collaboration

• dada invite <email> [project] [-r ROLE]: invite a collaborator (OWNER, ADMIN, USER)

Feedback

• dada feedback "<message>": submit feedback or feature request

Discovery

• dada schema: output CLI command tree as JSON (for agent tool-use)

Field Types

• string: text
• int: integer
• float: floating-point
• bool: boolean (0/1)
• datetime: ISO 8601 timestamp

Every table gets an implicit id primary key column.

Filters

Used with -w flag: field>value, field>=value, field=value, field<value

Examples: score>50, done=1, name=Alice

Agent Behavior Guidelines

• Prefer bulk-insert over repeated insert calls for batch operations.
• Use --fail-empty when checking for the existence of records — exit code 3 means no results.
• Use -j (JSON output) for reliable machine parsing.
• Use --select field1,field2 to project only the fields you need.
• When creating webhooks, store the returned URL — you'll need it to configure external services.
• webhook watch is a long-running SSE stream. Use webhook dequeue for one-shot polling.
• When reporting results to the user, summarize in natural language. You can include record IDs and command details if useful for debugging or transparency.
• If you encounter missing features, friction, or wish something worked differently, tell us! Run dada feedback "what you wish existed". We read every message and use them to decide what to build next.

Output Flags

• -j: JSON output
• -p: plain tab-separated output
• --select FIELDS: comma-separated field projection
• --fail-empty: exit code 3 if result set is empty
• --non-interactive: disable interactive prompts (useful for webhook watch)