David F GliddenandClaude Opus 4.8
3f9a89b00c
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>
2.6 KiB
name, description, metadata, permalink
| name | description | metadata | permalink | ||||||
|---|---|---|---|---|---|---|---|---|---|
| feedback-querying-default-palace-when-question-belongs-to-specific-palace | When the steward asks about the typography masters (or any subject whose canonical corpus lives in a non-default palace), default to checking which palace holds the answer before running queries. The MCP is bound to one palace; the corpus may live in another. |
|
claude-memory/feedback-querying-default-palace-when-question-belongs-to-specific-palace |
Querying the default palace when the question belongs to a specific palace
Rule: When the steward asks about content that lives in a known specialised palace (e.g., the typography masters, the chamber-library corpus), reach for the CLI against the specific palace first, not the MCP-bound default palace. If you don't know which palace holds the answer, ask before querying.
Why: The MCP (mcp__mempalace__* tools) is bound to a single palace at startup via ~/.mempalace/config.json — currently palace-memory (session continuity, claude-code KG, diaries). Specialised corpora live in separate palaces — currently palace-chamber-typography (40,894 drawers; Bringhurst, Tschichold, Hochuli, Lupton, Lacroux, et al.). Querying the MCP for a Bringhurst-shaped question returns conversation transcripts about Bringhurst, not Bringhurst's own prose — the wrong room. On 2026-05-20 the steward had to interrupt mid-search with "the specific typography palace, not the default" to redirect; the substantive answer to an em-dash question (in lists) was waiting in the chamber-typography palace the whole time.
How to apply: Before running mempalace queries on a corpus-shaped question, ask: which palace would hold this? If it's typography/masters → CLI against palace-chamber-typography. If it's session/diary/KG → MCP. If unsure → check ~/.mempalace/ for which palace directories exist, or ask the steward. See [[reference-mempalace-typography-palace]] for the canonical procedure.
This drift is adjacent to but distinct from acting-on-inferred-not-read (which is about claiming-without-verifying) and wrong-comparison-baseline-treated-as-current (which is about checking against the wrong git branch). The shared shape: the right answer is somewhere; reaching the wrong place by default. Discipline: name the source-of-truth before querying it.
KG entry: claude-code drift-pattern querying-default-palace-when-question-belongs-to-specific-palace (valid_from 2026-05-20). To be added to KG at /wrap-up.