菜单
Sweep Claude's private per-machine memory for team-relevant operational knowledge and propose moving it into the shared repo (or the private ops repo) as draft PRs. Use to promote lessons that accreted privately, or run on a schedule. Never auto-merges; never leaks.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Run OCR and translation quality evaluations across scripts and languages. Produces research-grade reports with MCR, cross-model agreement, embedding-space hallucination detection, and corpus readiness scores.
Run a Glass Bead Game session — a parlor game played against the Source Library corpus. Players connect verified passages from different books, times, and traditions into a growing string of 'beads,' with Claude as librarian, verifier, and scribe. Trigger when the user says 'glass bead game,' 'let's play beads,' or asks to start/resume/publish a game session. Claude runs the search-verify-display loop; humans do the noticing, choosing, and judging.
Stage-2 verification of first-translation flips using INDEPENDENT Claude subagents (subscription, not API key) with full evidence capture. Use to verify proposed badge changes (demotes/promotes) before writing any public claim — the binding step that catches AI-fabricated priors. Each subagent reports its actual search queries + sources, persisted to the first_translation_attempts provenance log.
Triage and respond to the user-feedback queue — pull open feedback, detect items already fixed but never marked, route translation-requests and metadata-corrections, mark items resolved, and (on approval) email submitters. Use when asked to check/triage/answer/clear feedback, "what are people saying?", or "respond to feedback."
One-shot audience & usage report — signups, DAU/MAU, time on-site, traffic, top content, search behavior, referrers, countries, and user journeys. Use when asked "what's our usage?", "how many signups?", "time on site?", "traffic?", or any audience/engagement metrics question.
Use this skill whenever choosing or evaluating a background image for the 'quote' section of a Source Library collection page — the wide banner where a sourced quotation is overlaid on an image drawn from the collection's plates or illustrations. Trigger for any request to pick, rank, or judge a background image behind quote text, select a plate for the quote banner, or decide whether a candidate image works as a text background. The goal is a beautiful, legible, on-theme backdrop that the quotation reads cleanly over.
| name | promote-lessons |
| description | Sweep Claude's private per-machine memory for team-relevant operational knowledge and propose moving it into the shared repo (or the private ops repo) as draft PRs. Use to promote lessons that accreted privately, or run on a schedule. Never auto-merges; never leaks. |
Operational know-how accretes in Claude's private, per-machine auto-memory
(~/.claude/projects/<project>/memory/, gitignored). It only reaches teammates
when someone notices and copies it into the shared repo. This skill makes that a
systematic, proposing sweep instead of a reactive one.
It runs weekly via the "Weekly knowledge-promotion sweep" routine, and you can
invoke it any time with /promote-lessons.
The test for repo memory vs private memory is:
Would a teammate's Claude need this to avoid taking a wrong action?
If yes → it belongs in the shared repo. The pause state, the
selective-unpause-strands-hidden-books gotcha, and real per-book cost all pass.
Derek's preferences, the API-key map, and in-flight scratch do not. Bias
operational facts into repo memory/ at /lesson time so there's less to sweep.
Read the private store. Open the index at
~/.claude/projects/<project>/memory/MEMORY.md (for this project the path is
~/.claude/projects/-Users-dereklomas-sourcelibrary/memory/MEMORY.md) and
skim the topic files it points to.
Classify each entry into one of three buckets:
memory/*.md or .claude/docs/.~/sourcelibrary-ops ONLY. Never the public repo.Dedupe against what's already shared. For each TEAM-TECHNICAL candidate,
grep memory/*.md and .claude/docs/*.md. Skip facts already represented.
Draft the additions. For genuinely-missing TEAM-TECHNICAL facts, add a
concise entry to the most relevant existing memory/ file, or a new
.claude/docs/ file when it's a standalone runbook. Group related facts into
one PR. For missing OPS/BUSINESS facts, add a note in ~/sourcelibrary-ops.
Open DRAFT PRs. Never merge. Never deploy. Derek reviews and merges.
This repo is public (AGPL). Never put secrets, API keys, PII, contacts, donor/fundraising info, or business strategy into it. When unsure whether something is shareable, route it to the private ops repo or skip it and flag it for review. A false "leave private" is cheap; a leak is not.
main. Do all edits in a worktree:
git worktree add -b <branch> .claude/worktrees/<name> main, then copy
src/lib/vendor/lamejs-bundle.js into the worktree's src/lib/vendor/ before
the first commit (the pre-commit check-imports hook needs it).Signed-off-by: JDerekLomas <j.d.lomas@tudelft.nl> (DCO).gh pr create --draft.Post a short summary: what was promoted (with draft-PR links), what was routed to ops, and what was left private and why. If nothing qualifies, say so in one line — a quiet week is a valid outcome.