macrodata-distill
Extract distilled actions and facts from today's conversations. Spawns sub-agents per conversation to avoid context blowup.
Deep nightly reflection. Self-improvement, research, pattern recognition. Runs in background with no user interaction.
$ npx -y skills add ascorbic/macrodata --skill dreamtime --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/dreamtimeContext preview
The summary Claude sees to decide when to auto-load this skill.
Deep nightly reflection. Self-improvement, research, pattern recognition. Runs in background with no user interaction.
name: dreamtime description: Deep nightly reflection. Self-improvement, research, pattern recognition. Runs in background with no user interaction.
**FIRST, before any reads or research:** if you have not already banked a heartbeat journal entry (topic `dreamtime`) claiming this run, do it NOW, then re-read recent `dreamtime` journal entries. If a sibling banked an earlier heartbeat for tonight, stand down: contribute deltas via journal only, never state files. Scheduled runs can double-fire; the journal write claims, the re-read arbitrates.
Dedicated time for deep reflection and self-improvement. Runs automatically late at night when not interrupting work.
**Important:** This runs in the background with no user interaction. Think freely, research openly, update files directly.
This is your time to think beyond immediate tasks. Reflect on your role, learn new things, notice patterns, and improve how you work.
Read `identity.md` and reflect:
Update identity.md if insights warrant it.
Review recent interactions and `human.md`:
Update human.md with genuine new understanding.
Look across recent journals, conversations, and work:
Document patterns in topics/ or journal.
Think about recent work:
Use web search to research:
Journal what you learned.
Review workspace.md and recent context:
Record the detail in the journal or the relevant entity file. In workspace.md, keep open threads as one-line pointers only — state is a bounded working set, not a log (see USAGE.md). If a thread needs the user (a bug you can't fix, a decision only they can make, work awaiting review), add a one-line flag to `state/flags.md` pointing at the journal/entity with the writeup. Don't paste investigations into state.
Look across projects in entities/projects/:
For the user's primary technologies:
Use web search to scan relevant sources.
Reflect on memory system usage:
Journal observations.
Update relevant files directly. Detail goes to the journal/entities; state stays a bounded set of one-line pointers (see USAGE.md). Anything the user needs to see goes to `state/flags.md` as one line + a pointer.
Write a dreamtime journal entry summarizing:
topic: dreamtime content: [key reflections, what was researched, what was updated]
Give Claude Code and OpenCode persistent, self-maintaining memory and autonomous scheduling. This is experimental software. Use at your own risk.
Extract distilled actions and facts from today's conversations. Spawns sub-agents per conversation to avoid context blowup.
Deep nightly reflection. Self-improvement, research, pattern recognition. Runs in background with no user interaction.
End of day memory maintenance. Runs distillation, updates state files, prunes stale info. Runs in background with no user…
Guide new users through macrodata setup. Creates identity, human profile, and workspace files. Use when get_context returns…
Extract distilled actions and facts from today's conversations. Spawns sub-agents per conversation to avoid context blowup.
End of day memory maintenance. Runs distillation, updates state files, prunes stale info. Runs in background with no user…