菜单
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.
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."
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.
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 | metrics |
| description | 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. |
Runs the two analytics scripts that together give the full audience + usage
picture, then summarizes the numbers and surfaces qualitative texture. Both read
production Mongo (bookstore) — purely read-only, no cost, no permission needed.
Live dashboard: a daily-refreshed version of these numbers (with trends, deltas, growth/DAU charts, and a methodology panel) lives at https://sourcelibrary.org/platform/admin/metrics (super-admin gated). It reads
system_config.metrics_snapshot, written byscripts/analytics/snapshot-metrics.mjson the Hetzner cron (05:45 UTC) and accumulated daily intometrics_history. Use the scripts below when you need fresher-than-daily numbers, a custom window (--days N), or the deeper qualitative texture (search chains, journeys) the dashboard intentionally omits.
set -a; source .env.production.local; set +a
node scripts/analytics/audience-metrics.mjs # signups, DAU/MAU, dwell, nudge list
node scripts/analytics/usage-deepdive.mjs --days 30 # traffic, content, search, journeys
node scripts/analytics/engagement-metrics.mjs # retention, conversion, reading depth, demand, social, AI, cost
usage-deepdive.mjs also writes a dated markdown report to
scripts/output/analytics/. All three take --days N (default 30) except
reading-depth in engagement-metrics, which is fixed at 7d for query speed.
audience-metrics.mjs — the numbers people ask for first:
usage-deepdive.mjs — the context:
engagement-metrics.mjs — the dimensions the other two leave out:
not_found_reports URLs = what people want and we lack (broken BPH/legacy links + genuinely missing pages)ai_usage by feature (librarian / search-expand / voice + cost + latency), api_usage by route/identity (the mcp route = external agents), distinct active API keysgemini_usage_daily spend (⚠ the daily rollup goes stale when the pipeline is paused — the script flags it; raw gemini_usage + ai_usage stay current)Lead with the headline numbers (signups + growth, MAU/DAU, time on-site), then
traffic and top content. Always add qualitative texture — the richest signal
is in usage-deepdive's search refinement chains (you can read individual
sessions: a HEMA scholar hopping through Fechtbücher → Talhoffer → Liechtenauer;
a Gnostic seeker hitting zero-results on bardo / Iamblichus / Gnostic ascent =
unmet acquisition need).
Caveats to state, not hide:
to of one = from of the next ARE one
ip_hash's real sequence, but most of the table stitches unrelated users. Do
NOT narrate it as one person's trail without pulling that ip_hash's raw rows
from search_queries first (sort by ts). to_n is the query's result count,
and a lone 0 is usually one anomalous search (e.g. Corpus Hermeticum medians
~13 results), not a content gap. Verify before storytelling.parace, fludd, th)
logged mid-keystroke — not real content gaps. Only conceptual zero-results
(bardo, Gnostic ascent) point at genuine missing content.