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>
4.5 KiB
name, description, type, originSessionId, permalink
| name | description | type | originSessionId | permalink |
|---|---|---|---|---|
| Chavruta skill — parked, expected to be load-bearing for ARC editorial work | A `/chavruta <topic>` skill that maps a topic to a known voice slate, reads source markdowns directly (bypassing mempalace search), returns convergence/divergence. Steward called for it 2026-05-06 after mempalace search returned Chicago TOC junk over the actual typographer voices. Build when ARC work has a quiet hour. Steward expects to "be calling on it quite a bit." | project | 3bf7ae75-1138-4745-9024-0c40b3755ad3 | claude-memory/project-chavruta-skill-parked-2026-05-06 |
Chavruta skill — parked 2026-05-06
Origin
Surfaced 2026-05-06 during Sources-styling decision on Death of a Hero. MemPalace search returned 8 duplicate copies of Chicago Manual TOC link-junk while the actual Bringhurst §5.2.4, Morison BIBLIOGRAPHY AND REFERENCES, and Rutter "outdent in bibliography" passages — all mined — failed to surface. Direct grep against source markdowns found them in seconds.
Steward (verbatim 2026-05-06): "keep the idea close, we'll build it once we've cleared some more of the ARC todos. I think that we will be calling on it quite a bit to be honest"
Why a skill, not an agent
The pattern is read-known-sources-and-synthesise — light, fast, deterministic. Skills are the right register. Agent infrastructure would be overkill.
Why this beats mempalace search for ARC editorial work
The chavruta protocol assumes a known voice slate — Bringhurst+Morison+Hochuli+Rutter for typography; Vico+Leopardi+Harrison+Bachelard for phenomenology; etc. MemPalace tries to replace that slate-knowledge with semantic similarity and currently fails (duplicate-chunks issue + mxbai 384d retrieval weakness). The skill makes the steward's existing slate-knowledge structural rather than ad-hoc.
It also bypasses mempalace's current degradation entirely. While the substrate is still being rebuilt/improved, this skill provides reliable consultation.
Sketch of structure
~/.claude/skills/chavruta/SKILL.md— protocol description, when to invoke~/.claude/skills/chavruta/voice-slates.md— domain → voice slate mapping (typography, phenomenology, philosophy of education, etc.)- Invocation:
/chavruta <topic>or/chavruta <topic> --slate typography - Procedure: select slate → grep relevant chapters in chamber-library → return convergence + divergence frame
- Output shape: like today's typographer chavruta on Sources styling — direct quotes, attribution, named convergence + divergence, optional editorial recommendation
Voice slate candidates (initial — to refine when building)
- Typography: Bringhurst (Elements, Solid Form) + Hochuli (Detail) + Morison (First Principles) + Rutter (Web Typography) + Tschichold + Hart's Rules
- Bibliography & editorial conventions: Bringhurst + Harrison's Dominion of the Dead Works Cited + Hart's Rules (European register) + Norris (Between You & Me) for English usage
- Phenomenology / dwelling / lares: Bachelard (Poetics of Space, Reverie) + Heidegger + Vico (Scienza Nuova) + Harrison (Dominion, Forests, Juvenescence)
- Music & performance: voices to gather (Harnoncourt, Taruskin, Heartz, Swafford, Norman, Leopold Mozart…)
- Pattern / craft / inhabitation: Alexander (Timeless Way, Pattern Language, Oregon, Nature of Order) + Sennett (Craftsman) + Mumford (Technics)
Expected uses
ARC editorial decisions across:
- Typographic/apparatus questions (footnotes, sidenotes, bibliographies, ornament register, etc.)
- Voice convergence in essay drafting (e.g., "what do my phenomenology voices say about
dwelling?") - Style-guide build-out (the lex-method work)
- Memento Mori print decisions when that work begins (it will need different slates than web)
Steward's signal that this will be high-frequency means the slate-mapping needs to be carefully composed at build time, not improvised. Worth ~half a day's careful work to set up well rather than ~1 hour to set up sketchily.
Build trigger
When ARC todos have a clearer queue (post-Chamber archive landing? after a major composition arc closes?). Don't build mid-task. Don't build under fatigue.
Cross-references
- 2026-05-06 session — origin context
upstream-issue-duplicate-chunks-2026-05-06.md— the mempalace problem this skill works aroundproject-mempalace-role-as-instrumentation.md— mempalace as instrumental, not permanent — chavruta skill is a parallel instrumental tool