用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/smithersai/smithers --skill smithers-workflow命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Drive Smithers, a durable control plane for long-running coding agents, from Claude Code. Use when the user wants multi-step, long-running, crash-safe, or human-in-the-loop agent work ('orchestrate agents', 'run a workflow', 'implement this and review it', 'keep iterating until tests pass', 'plan then build') or anything needing retries, approvals, replay, or evals across multiple AI steps. YOU (Claude) run Smithers on the user's behalf; it is not a GUI the human clicks. HARD RULE 0, if `SMITHERS_INSIDE_RUN` is set in your environment you are ALREADY a worker agent inside a Smithers node, so never invoke the Smithers CLI or MCP tools to launch or steer a run, just do the node's task directly with your ordinary tools; every rule below applies only when it is unset. HARD RULE 1, right-size the route FIRST, handle ANY clear single-goal ask directly, even a large repo-wide one, and reserve a full workflow for work that genuinely needs ordered stages, durability, approvals, loops, or reuse; never author a multi-no
Drive Smithers, a durable control plane for long-running coding agents, from Codex. Use when the user wants multi-step, long-running, crash-safe, or human-in-the-loop agent work ("orchestrate agents", "run a workflow", "implement this and review it", "keep iterating until tests pass", "plan then build") or anything needing retries, approvals, replay, or evals across multiple AI steps. YOU (Codex) run Smithers on the user's behalf; it is not a GUI the human clicks. HARD RULE 0: if `SMITHERS_INSIDE_RUN` is set in your environment you are ALREADY a worker agent inside a Smithers node — never invoke the Smithers CLI or MCP tools to launch or steer a run, just do the node's task directly with your ordinary tools; every rule below applies only when it is unset. HARD RULE 1: right-size the route FIRST, handle a clear well-scoped single-goal task directly, and reserve a full workflow for work that genuinely needs ordered stages, durability, approvals, loops, or reuse. A workflow that runs long, fans out, or pauses on
Drive Smithers, a durable control plane for long-running coding agents. Use when the user wants multi-step, long-running, crash-safe, or human-in-the-loop agent work: "orchestrate agents", "run a workflow", "implement this and review it", "keep iterating until tests pass", "plan then build", or anything that needs retries, approvals, replay, or evals across multiple AI steps. YOU (the agent) run Smithers on the user's behalf; it is not a GUI the human clicks. HARD RULE 0: if `SMITHERS_INSIDE_RUN` is set in your environment you are ALREADY a worker agent inside a Smithers node, so never invoke the Smithers CLI or MCP tools to launch or steer a run; do the node's task directly with your ordinary tools. Everything below applies only when it is unset. You are otherwise an ORCHESTRATOR: run long-running, multi-step, or background work *through* Smithers, not through your own ad-hoc subagents; spend your time observing the run and reporting. Also governs DELIVERABLE SHAPE even when no workflow is involved: a user
基于 SOC 职业分类
正在显示 SKILL.md
| name | smithers-workflow |
| description | Discover local workflows from .smithers/workflows. Run `smithers workflow --help` for usage details. |
| requires_bin | smithers |
| command | smithers workflow |
Create a new flat workflow scaffold in .smithers/workflows (or the global ~/.smithers with --global).
| Name | Type | Required | Description |
|---|---|---|---|
name | string | yes | Workflow ID |
| Flag | Type | Default | Description |
|---|---|---|---|
--global | boolean | false | Create the workflow in the global ~/.smithers pack (honors SMITHERS_HOME) instead of the local .smithers |
Inspect workflow discovery, preload files, and detected agents.
| Name | Type | Required | Description |
|---|---|---|---|
name | string | no | Workflow ID |
Show workflow metadata and an agent-facing skill preview.
| Name | Type | Required | Description |
|---|---|---|---|
name | string | yes | Workflow ID |
List discovered local workflows. System workflows are hidden unless --system is passed.
| Flag | Type | Default | Description |
|---|---|---|---|
--system | boolean | false | Include system workflows (internal plumbing hidden from the default listing) |
Resolve a workflow ID to its entry file.
| Name | Type | Required | Description |
|---|---|---|---|
name | string | yes | Workflow ID |
Run a discovered workflow by ID. Omit the ID (or pass --interactive) to pick one interactively.
| Name | Type | Required | Description |
|---|---|---|---|
name | string | no | Workflow ID (omit with --interactive to pick one) |
| Flag | Type | Default | Description |
|---|---|---|---|
--detach | boolean | false | Run in background, print run ID, exit |
--runId | string | Explicit run ID | |
--parentRunId | string | Existing run ID to record as this run's parent (persisted lineage, surfaced by inspect/ps and the MCP run tools) | |
--maxConcurrency | number | Maximum parallel tasks (default: 4) | |
--root | string | Tool sandbox root directory | |
--log | boolean | true | Enable NDJSON event log file output |
--logDir | string | NDJSON event logs directory | |
--allowNetwork | boolean | false | Allow bash tool network requests |
--maxOutputBytes | number | Max bytes a single tool call can return | |
--toolTimeoutMs | number | Max wall-clock time per tool call in ms | |
--hot | boolean | false | Enable hot module replacement for .tsx workflows |
--input | string | Input data as JSON string | |
--annotations | string | Run annotations as a flat JSON object of string/number/boolean values | |
--resume | unknown |
Generate agent-facing skill docs for local workflows.
| Name | Type | Required | Description |
|---|---|---|---|
name | string | no | Workflow ID, or omit to generate skills for all workflows |
| Flag | Type | Default | Description |
|---|---|---|---|
--output | string | Output file for one workflow, or output directory for all workflows | |
--force | boolean | false | Overwrite existing skill files |
--global | boolean | false | Write skills into the global ~/.smithers pack (honors SMITHERS_HOME) instead of the local .smithers |
false |
| Resume a previous run. Pass true with --run-id, or pass the run ID directly (e.g. --resume ) |
--force | boolean | false | Resume even if still marked running |
--resumeClaimOwner | string | Internal durable resume claim owner |
--resumeClaimHeartbeat | number | Internal durable resume claim heartbeat |
--resumeRestoreOwner | string | Internal durable resume restore owner |
--resumeRestoreHeartbeat | number | Internal durable resume restore heartbeat |
--serve | boolean | false | Start an HTTP server alongside the workflow |
--supervise | boolean | false | Run the stale-run supervisor loop (with --serve) |
--superviseDryRun | boolean | false | With --supervise, detect stale runs without resuming |
--superviseInterval | string | 10s | With --supervise, poll interval (e.g. 10s, 30s) |
--superviseStaleThreshold | string | 30s | With --supervise, stale heartbeat threshold |
--superviseMaxConcurrent | number | 3 | With --supervise, max runs resumed per poll |
--port | number | 7331 | HTTP server port (with --serve) |
--host | string | 127.0.0.1 | HTTP server bind address (with --serve) |
--authToken | string | Bearer token for HTTP auth (or set SMITHERS_API_KEY); required to bind a non-loopback --host |
--insecure | boolean | false | Allow binding a non-loopback --host with NO auth (exposes unauthenticated approve/deny/cancel control of the run — dangerous) |
--metrics | boolean | true | Expose /metrics endpoint (with --serve) |
--backend | string | Bootstrap storage selection for a workflow owner or workspace Gateway; not a run-discovery/control flag |
--postFailure | boolean | true | Auto-launch the post-failure autopsy workflow when this run fails (disable with --no-post-failure or SMITHERS_POST_FAILURE=0) |
--verbose | boolean | false | Show engine info logs (run lifecycle, agent sessions) on interactive runs; the default keeps progress lines + warnings only. Non-TTY/structured output always gets full logs. |
--report | boolean | true | On an interactive run, narrate the result with a cheap/fast agent and open an HTML summary in the browser when it finishes (disable with --no-report or SMITHERS_NO_REPORT=1). |
--prompt | string | Prompt text mapped to input.prompt when --input is omitted |
--interactive | boolean | false | Pick a workflow and its inputs through interactive terminal prompts, then launch the full-screen TUI monitor for the run (TTY only) |
