From kagura-memory
Summarizes the current development session and saves key learnings, decisions, patterns, and bug fixes to Kagura Memory Cloud for recall in future sessions.
How this command is triggered — by the user, by Claude, or both
Slash command
/kagura-memory:session-summaryThe summary Claude sees in its command listing — used to decide when to auto-load this command
Summarize the current session's key learnings and save them to Kagura Memory Cloud. ## When to use At the end of a development session, or when switching to a different task. Captures decisions, patterns, bugs, and plans that would be useful in future sessions. ## Steps ### 1. Identify related GitHub issues Review the conversation and collect all GitHub issue numbers that were worked on, referenced, or discussed. Use `gh issue view <number>` if needed to confirm titles. ### 2. Identify what to remember Review the conversation and categorize knowledge into one of the canonical types (...
Summarize the current session's key learnings and save them to Kagura Memory Cloud.
At the end of a development session, or when switching to a different task. Captures decisions, patterns, bugs, and plans that would be useful in future sessions.
Review the conversation and collect all GitHub issue numbers that were worked on, referenced, or discussed. Use gh issue view <number> if needed to confirm titles.
Review the conversation and categorize knowledge into one of the canonical types (see remember skill for the full vocabulary):
| Type | What to capture | Importance |
|---|---|---|
decision | Architecture decisions, approach choices, rejected alternatives with WHY | 0.8-1.0 |
pattern | Implementation patterns, technical solutions, reusable approaches | 0.7-0.9 |
bug-fix | Root cause analysis, regression lessons, "never do this again" | 0.8-0.9 |
troubleshooting | Error solutions, workarounds, environment-specific fixes | 0.5-0.7 |
learning | SDK benchmark results, performance findings, tool limitations | 0.6-0.8 |
note | Milestone status, roadmap changes, issue relationships | 0.6-0.8 |
time | Forward-looking / dated follow-ups (a deadline, "re-check on date X", a scheduled flag flip). Set details={"trigger": {...}} so recall_upcoming surfaces it at the right time instead of letting it decay into the backlog. | 0.6-0.8 |
list_contexts()
Ask the user which context to save to if unclear. Default: the project's development context.
For each item, use remember with:
category:{domain} + entity tags + writing variations for Japanese + issue:#N for each related issueRelated issues: #N, #M line at the end of content linking to relevant GitHub issues.["vault://my-vault/related-note.md"]). Unresolved URIs are silently skipped.delivery_mode="always") — sparinglyA pinned memory is loaded deterministically every session via load_pinned (see session-start). That makes it the right home for a true standing invariant — an active prod color, a non-negotiable policy, a release gate — but it also means every pinned memory costs context budget on every future session. Over-pinning degrades the signal of every session that follows.
Pin only when the knowledge must be seen every session, not merely recalled when relevant. Concretely:
pinned_load_cap, default 100) is a truncation safety net, not the editorial budget — do not let it stand in for judgment. Decisions, patterns, and status notes are NOT pin material; they belong in normal recall.load_pinned(context_id=...). If the context already has ≥7 pinned, unpin a stale one first.green → blue), unpin the old one in the same step — never leave two competing invariants pinned:
update_memory(memory_id=<old>, context_id=..., delivery_mode="on_recall")
Then remember(..., delivery_mode="always") (or update_memory(... delivery_mode="always")) the new one.To pin at save time: remember(..., delivery_mode="always").
type="troubleshooting" with tags=["dev-environment"] so future sessions can avoid the same mistakeAfter saving, print a summary:
## Session Knowledge Saved
Related issues: #12, #34
| # | Type | Summary | Importance |
|---|------|---------|------------|
| 1 | decision | ... | 0.9 |
| 2 | pattern | ... | 0.8 |
...
Total: N items saved to context {context_name}
npx claudepluginhub kagura-ai/memory-cloud --plugin kagura-memory