Basic Memory v0.21.6 first sync over the live memory dir (steward-authorized live-dir trial, Option A 2026-06-06): adds permalink: to frontmatter, refolds long YAML description lines, strips final newlines. Bodies untouched — verified via full diff classification. From this commit forward, any diff in claude/memory shows only what Basic Memory or the session writes. Trial design: MemPalace untouched as incumbent; git status check on this dir at every wrap; end-of-day evaluation (recall quality, sync robustness, rebuild-from-files, malformed-file behavior). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
1.4 KiB
name, description, metadata, permalink
| name | description | metadata | permalink | ||||||
|---|---|---|---|---|---|---|---|---|---|
| mempalace-kg-object-128-char-cap | MemPalace kg_add `object` field hard-caps at 128 chars — write KG objects short on the FIRST pass; put detail in the drawer. |
|
claude-memory/feedback-mempalace-kg-object-128-char-cap |
mcp__mempalace__mempalace_kg_add rejects any object longer than 128 characters ("object exceeds maximum length of 128 characters"). This recurs at nearly every /wrap-up §5 — I write a rich, sentence-length object, it fails, I re-add concise. The steward flagged it 2026-05-31: "this happens every time."
Why: the KG is a graph foothold for retrieval, not the record. The full session detail belongs in the drawer (add_drawer, no such cap) and the session memory file. The triple just needs to be findable by kg_query on the subject.
How to apply: at /wrap-up §5, compose every kg_add object as a short phrase under 128 chars on the first call — keyword-dense, not a sentence. Park the elaboration (commits, rationale, caveats) in the drawer + memory file, and if a pointer is useful, end the object with …; see <doc> §X. Subject/predicate also stay terse. One write, not two.
Sibling discipline: rank-on-fields-you-actually-write — verify tool field constraints before composing for them, don't re-derive at call-time.