Context
Mutating commands (create, install, remove) execute immediately with no way to preview what will happen. A --dry-run flag could show the planned actions without executing them.
Idea
Add --dry-run flag to skern skill create, skern skill install, skern skill remove, and other mutating commands.
Open Questions — Decisions Needed
The value and design of this feature need clarification:
-
Agentic use case is unclear — It's not obvious how agents would treat --dry-run output. Two potential scenarios:
- Agent → sub-agent handoff: An orchestrating agent uses
--dry-run to plan, then passes the plan to a sub-agent for execution. This requires structured (--json) dry-run output.
- Human review: A human uses
--dry-run to preview before confirming. This is a more traditional CLI pattern.
-
Is this redundant for agents? — Agents can already use skern skill search and skern skill show to inspect state before acting. Does --dry-run add enough value over the existing inspection commands to justify the implementation cost?
-
Output format — What does dry-run output look like?
- Same as regular output but with a "would do" prefix?
- A structured plan object in
--json mode?
- A diff-like format showing before/after state?
-
Scope — Which commands get --dry-run? All mutating commands, or only the risky ones (remove, uninstall)?
These questions should be resolved to determine whether this feature is worth pursuing and what form it should take.
References
Analysis item 2.9
Context
Mutating commands (
create,install,remove) execute immediately with no way to preview what will happen. A--dry-runflag could show the planned actions without executing them.Idea
Add
--dry-runflag toskern skill create,skern skill install,skern skill remove, and other mutating commands.Open Questions — Decisions Needed
The value and design of this feature need clarification:
Agentic use case is unclear — It's not obvious how agents would treat
--dry-runoutput. Two potential scenarios:--dry-runto plan, then passes the plan to a sub-agent for execution. This requires structured (--json) dry-run output.--dry-runto preview before confirming. This is a more traditional CLI pattern.Is this redundant for agents? — Agents can already use
skern skill searchandskern skill showto inspect state before acting. Does--dry-runadd enough value over the existing inspection commands to justify the implementation cost?Output format — What does dry-run output look like?
--jsonmode?Scope — Which commands get
--dry-run? All mutating commands, or only the risky ones (remove,uninstall)?These questions should be resolved to determine whether this feature is worth pursuing and what form it should take.
References
Analysis item 2.9