| name | hovel |
| description | Operate Hovel through its typed MCP interface. Use for Hovel workspaces, operations, chains, targets, modules, throws, payloads, sessions, or artifacts. |
| license | Apache-2.0 |
| compatibility | Requires Hovel 0.4.x and a configured Hovel MCP server. |
| metadata | {"hovel-min-version":"0.4.0","hovel-max-version":"0.5.0"} |
Hovel
Use Hovel as an auditable operator platform, not as a generic command shell.
The stable mental model is:
workspace
operation
targets
chains
modules
configuration
validation
throw
installed payloads
sessions
artifacts
Start with hovel_operator_identity and hovel_workspace_snapshot. Use
hovel_catalog_snapshot before choosing capabilities. Reuse suitable existing
state instead of creating duplicates.
Prefer typed Hovel MCP tools. Use hovel_command_run only when no typed tool
represents the capability. Never treat validation as authorization to execute.
Load the specialized Hovel skill matching the task:
- discovery and state orientation:
hovel-discovery
- operations, targets, chains, modules, and validation:
hovel-chain-building
- planning, confirming, and starting a throw:
hovel-throw
- installed payload operations:
hovel-payloads
- established session interaction:
hovel-sessions
- result and artifact inspection:
hovel-artifacts
- failures, stale state, and recovery:
hovel-troubleshooting
Do not conflate a target with an installed payload, a payload with a session,
an artifact with a payload, a chain with a throw, or a valid plan with approval.
Read references/workflow.md and references/safety-model.md when the task
crosses more than one lifecycle boundary.
