skillfile
CLI for the Skillscript agent workflow runtime
TLDR
Install runtime, init project, open dashboard
$ skillfile init
skillfile dashboard```
**Lint**, **compile**, and **execute** a skill
skillfile compile [skill]skillfile execute [skill]```Approve a draft skill (secured mode)skillfile dashboard```
**Lint**, **compile**, and **execute** a skill
$ skillfile approve [hello]
Show health, diagram, or replay a trace$ skillfile diagram [skill]
skillfile replay [trace_id]```
# SYNOPSIS
**skillfile** _command_ [_options_] [_args_]
# DESCRIPTION
**skillfile** is the command-line interface for **Skillscript**, a constrained declarative language and runtime for agent-authored, human-approved automation. Skills look like Make-inspired workflows with variables, connectors, and allowlists—without unrestricted shell or arbitrary package installs.
The runtime (**skillscript-runtime** on npm) executes skills via configured connectors, filesystem and shell allowlists, and optional secured mode with operator signatures. Agents typically author skills over MCP (**http://localhost:7878/rpc** when the dashboard is up); operators review, approve, and run them with **skillfile**.
# PARAMETERS
**init**
> Initialize a local Skillscript project
**dashboard** [**--host**] [**--port**]
> Start the operator UI and MCP endpoint (default port **7878**)
**lint** / **compile** / **execute** _skill_
> Validate, compile, or run a skill
**approve** _skill_
> Approve a draft skill for execution
**diagram** / **fires** / **replay** / **health**
> Visualize, inspect triggers, replay traces, or check runtime health
# CONFIGURATION
Environment controls include **SKILLSCRIPT_SHELL_ALLOWLIST**, **SKILLSCRIPT_FS_ALLOWLIST** (default deny), **SKILLSCRIPT_SECURED_MODE**, and **SKILLSCRIPT_SECRET_*** for runtime-held secrets. Project config examples ship as **skillscript.config.json.example** and **connectors.json.example**.
# CAVEATS
Pre-1.0; language and connector contracts are still stabilizing. Shell and filesystem access are deny-by-default until allowlisted. Bind the dashboard to **0.0.0.0** only with appropriate network protection.
# SEE ALSO
make(1)
# RESOURCES
skillfile replay [trace_id]```
# SYNOPSIS
**skillfile** _command_ [_options_] [_args_]
# DESCRIPTION
**skillfile** is the command-line interface for **Skillscript**, a constrained declarative language and runtime for agent-authored, human-approved automation. Skills look like Make-inspired workflows with variables, connectors, and allowlists—without unrestricted shell or arbitrary package installs.
The runtime (**skillscript-runtime** on npm) executes skills via configured connectors, filesystem and shell allowlists, and optional secured mode with operator signatures. Agents typically author skills over MCP (**http://localhost:7878/rpc** when the dashboard is up); operators review, approve, and run them with **skillfile**.
# PARAMETERS
**init**
> Initialize a local Skillscript project
**dashboard** [**--host**] [**--port**]
> Start the operator UI and MCP endpoint (default port **7878**)
**lint** / **compile** / **execute** _skill_
> Validate, compile, or run a skill
**approve** _skill_
> Approve a draft skill for execution
**diagram** / **fires** / **replay** / **health**
> Visualize, inspect triggers, replay traces, or check runtime health
# CONFIGURATION
Environment controls include **SKILLSCRIPT_SHELL_ALLOWLIST**, **SKILLSCRIPT_FS_ALLOWLIST** (default deny), **SKILLSCRIPT_SECURED_MODE**, and **SKILLSCRIPT_SECRET_*** for runtime-held secrets. Project config examples ship as **skillscript.config.json.example** and **connectors.json.example**.
# CAVEATS
Pre-1.0; language and connector contracts are still stabilizing. Shell and filesystem access are deny-by-default until allowlisted. Bind the dashboard to **0.0.0.0** only with appropriate network protection.
# SEE ALSO
make(1)
# RESOURCES
$ Homepage