Stats
Actions
Tags
From opencode
How to surface output from `companion review`, `companion adversarial-review`, and `companion rescue` to the user. Use when handing back the stdout of a companion call.
How this skill is triggered — by the user, by Claude, or both
Slash command
/opencode:opencode-result-handlingThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
opencode's response to a review or rescue call is the user-facing answer. The
opencode's response to a review or rescue call is the user-facing answer. The
slash commands and the opencode:opencode-rescue subagent must return that
stdout without modification:
If the user wants follow-up action on a review finding, they will ask. The contract is: review describes, the user decides.
When there are no changes to review the companion prints a single line:
No changes to review.
Surface that line as-is. Do not try to manufacture a review prompt around a clean tree.
/opencode:review --background prints a job id banner and returns immediately:
opencode review started: <id>
Check progress with /opencode:status. Fetch output with /opencode:result.
The eventual review body lives at the path returned by /opencode:result. When
the user runs /opencode:result, return its stdout verbatim — same rule.
The companion writes errors to stderr and exits non-zero. Examples:
Not a git repository: <cwd>opencode CLI not found on PATH. Run /opencode:setup.Invalid --model value: <value>Surface these exactly. They contain the next action the user needs.
npx claudepluginhub joelfrancisco/opencode-plugin-cc --plugin opencodeCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.