← all publishers

stackbone

@stackbone source repo

4 published skills

  1. Stackbone · stackbone bundle
    Use this skill when writing the inside of a Stackbone workspace with @stackbone/sdk: the workspace is discovered by convention from the files on disk (deep-agents/<name>/index.ts + workflows/<name>.workflow.ts), authoring a deep agent (a single index.ts default-exporting defineDeepAgent({ model, systemPrompt, tools, subagents, interruptOn }) from @stackbone/sdk/deep, with LangChain tools and connectorTool(...) for third-party operations), writing durable workflows as 'use workflow' / 'use step' functions with sibling inputSchema/outputSchema, calling an agent from a workflow step with callDeepAgent(name, input) from @stackbone/sdk/workflow, and reaching the ambient `stackbone` client from inside a tool or a workflow step — stackbone.database (Drizzle over Neon Postgres), stackbone.storage (file uploads), stackbone.ai (LLM chat / embeddings via OpenRouter), stackbone.rag (ingest + hybrid search), stackbone.config / stackbone.secrets / stackbone.prompts, and stackbone.connection(id) to call a third-party connec
    0
    installs
  2. Stackbone CLI · stackbone bundle
    Use this skill when driving the `stackbone` CLI to scaffold, develop, publish or operate an agent on Stackbone: authenticating with device flow (login / whoami), scaffolding a workspace shell and linking it to the control plane (init, optionally with a first agent / workflow / workflow-agent piece), adding more pieces offline into an existing workspace (add deep-agent / add workflow / add workflow-agent), linking an existing directory to an agent (link), running the local dev stack with Studio (dev), publishing the workspace (publish), switching the active organization the CLI acts as (organization use), managing database migrations with Drizzle (db migrate up / create / status), reading workspace metadata in JSON (metadata), listing the available connectors and the account's connections with ids and names (connectors), printing inline docs (docs), and operating a running agent installation from the shell — discovering install targets (agents), inspecting and triggering the durable workflows it exposes (workf
    0
    installs
  3. Stackbone Coder · stackbone bundle
    Use this skill to GENERATE a new Stackbone piece from a clean idea through a guided interview: first decide what to build — a deep agent, a durable workflow, or a workflow-agent (a workflow that calls an agent) — then scaffold it with the stackbone CLI and interview the user surface by surface. For an agent you ask for its tools and its system prompt; for a workflow you ask for the input data and the output data (the inputSchema / outputSchema); then you walk the capability checklist ONE AT A TIME — do we need a database? human-in-the-loop? RAG? storage? an LLM call? a connector? prompts? config? secrets? a schedule? to call another agent? — and wire in only the surfaces the user actually needs. This skill DRIVES the interview and the scaffolding; it delegates the real code to the `stackbone` skill (SDK authoring) and the commands to the `stackbone-cli` skill. Trigger on requests like: build a new agent, scaffold a workflow, create a workflow that calls an agent, "I want to make an agent that …", "I need a pi
    0
    installs
  4. Stackbone Debug · stackbone
    Use this skill when triaging an error, failure or unexpected behavior in a Stackbone workspace — a deep agent or a durable workflow running locally under `stackbone dev`, or an install deployed to a cloud / self-host. Covers SDK errors ({ data, error } envelope codes from the catalog returned by the ambient `stackbone` client), HTTP 4xx/5xx from the control plane, durable run failures and timeouts, `workflow_input_invalid` (a workflow start whose input failed its declared schema), HITL stuck at either level (a workflow parked on a `requestApproval()` gate, or an agent turn `interrupted` on a tool gated with `interruptOn`), durable-session issues (memory not surviving because `x-stackbone-session` wasn't sent, or the checkpointer package is missing), database slow queries / missing indexes / pgvector mismatches, secrets decryption errors, connector action failures via `stackbone.connection(id)` (ambiguous / unauthorized connections), publish/build failures (SDK inlined, native dep), tier quota exceeded respons
    0
    installs