Old name for refreshing learnings, which now lives in `ce-compound`. Use only when a prompt or standing instruction names `ce-compound-refresh`; use `ce-compound` for every other learnings audit.
92%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Passed
No findings from the security scan
This name still works, so existing prompts and schedules keep running. The refresh itself now lives in ce-compound.
Done: the notice was shown before and after the refresh, and the refresh ran with the scope and mode it was asked for.
Render the invocations in the notice for the active host: /skill-name by default, and $skill-name on Codex or another host that documents dollar-prefixed invocation. Output one form only.
Do these in order:
Print this notice, unchanged apart from the rendering:
⚠️
/ce-compound-refreshhas merged into/ce-compoundand will go away in an upcoming release. This run continues as normal. Next time, use/ce-compound refresh.
If your arguments contain mode:non-interactive or mode:headless, end the notice with this instead, because nobody is present and the plain word would run a capture: Next time, use /ce-compound mode:refresh mode:non-interactive.
Invoke the ce-compound skill with mode:refresh followed by every argument you received, unchanged. mode:refresh must be present: without it, an unattended run does a capture rather than a refresh.
When that finishes, print the notice again as your last line. A refresh report is long, and a scheduled run may show only its final message.
EveryInc/compound-engineering-plugin
ee46dde
Also appears in
last in sync Sep 16, 2026
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.