Flush

Skill kevin-hs-sohn/hipocampus/skills/flush

Manual memory flush: dump current session context to daily raw log via subagent. Invoke with /hipocampus:flush. Run hipocampus:compaction afterwards for tree propagation and qmd reindex.From its SKILL.md

Install

npx -y skills add kevin-hs-sohn/hipocampus --skill flush

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

1.4 KB, as published. Nobody here has run it

Hipocampus Memory Flush

Dump current session context to the daily raw log. Use when you want to persist what happened in this session without waiting for End-of-Task Checkpoint or context compression.

For full compaction (needs-summarization processing, tree propagation, qmd reindex), run hipocampus:compaction skill after this.

Steps

1. Compose session summary

Gather a summary of everything discussed in this session so far. For each topic:

## [Topic Name]
- request: what the user asked
- analysis: what you researched/analyzed
- decisions: choices made with rationale
- outcome: what was done, files changed
- references: knowledge/ files, external sources

2. Dispatch subagent

Dispatch a subagent with the session summary and this task:

Hipocampus memory flush. Append the following structured log to memory/YYYY-MM-DD.md (today's date). Then run hipocampus compact to propagate through the tree.

[paste session summary here]

The subagent writes the files and runs compact. The main session stays clean.

3. Report

Confirm the flush completed:

Topics flushed
Subagent status