David/dotfiles
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

session 2026-06-16 (eve): pattern-finder built + Station-I ingested + pass-1 contamination caught + ARC ornaments grounded

Browse Source
This commit is contained in:
David F Glidden
2026-06-16 21:44:56 +02:00
parent 555d0c2eec
commit 5af1b4d1ad
4 changed files with 89 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
claude/memory/skill-harvest-register.md
+14
View File
@@ -186,3 +186,17 @@ The single place proposed skills live so they don't evaporate between sessions.
- Positive language-validity check (dictionary-valid-word ratio per language) for verify_conversion — catches "no cruft fires but text is OCR garbage."
- Promote today's catalogue census (title+author co-required, evidence-emitting) into a reusable `census_membership` tool.
- Calibration follow-ups logged: word_sanity floor vs genuine short fragments; repair coverage threshold should weight content over front-matter.
**New proposal (2026-06-16 wrap §1.6 — awaiting steward):**
| Skill | Kind | One-line | Status |
|---|---|---|---|
| `/wrap-up` §4.b/§5 | patch | Inline reminder at the KG-write step: `kg_add` `object` hard-caps at **128 chars** — write short keyword objects on the FIRST pass (detail in the drawer). Hit it AGAIN this wrap (2 failed adds, re-added short). The memory [[feedback-mempalace-kg-object-128-char-cap]] knows it; the skill doesn't carry the reminder at the moment of writing. A one-line patch would stop the recurrence. | PROPOSED |
### Harvest 2026-06-16 (evening — pattern-finder build + Station-I contamination)
| Element | Kind | One-line | Where it lands | Status |
|---|---|---|---|---|
| **Work-in-omnibus: verify the interior, not just the endpoints** | verification-ladder instrument | When a sidecar/scope brackets one work out of a **multi-work source** by heading-to-heading boundaries, **content-sample the span INTERIOR** (mid-span reads) and scan for embedded works / editorial apparatus **before trusting the bracket** — endpoint-verification is necessary but not sufficient. Caught the hard way this session: my La Chute bracket `[8062–10951]` verified at both ends (Clamence present, Meursault absent) yet its interior was *mostly Le Premier Homme* — a whole undeclared novel the heading-to-heading logic swallowed. Specialization of [[trust-prior-pass-frame]] (verify at the scope of the extension), kin to the censused-routes flag. Lands on [[reference-verification-ladder]]. | verification ladder | **PROPOSED** |
| **studium-engine tool-evolution-log** | create | Establish the analog of `chamber-library/_curation/tool-evolution-log.md` for the **engine** tools (pattern_finder, ingest_gate, chunker, store, retrieve), per the standing [[feedback-tool-review-after-each-use]] discipline — studium tool-reviews currently have no home. First entry = the pattern_finder pass-1 review (**PASS-BUT-FALSELY**: reported "5/5 voices" over partly-false grounding). Home candidate: `studium-engine/docs/_audits/tool-evolution-log.md`. | studium-engine repo | **PROPOSED** |
*(Both load-bearing: the interior-homogeneity instrument would have prevented the Camus contamination outright; the engine tool-log is where this very review should have been recorded and wasn't. Neither created autonomously — surfaced for steward authorization.)*