Weekend Tech Vocab Scheduler
Documents the recurring cloud routine that keeps tech_english_vocab growing automatically. This is a Claude Code cloud routine (Anthropic-hosted CCR session), not part of the local Hermes nightly cron chain (hermes-harness-spec) — it's a separate, isolated sandbox with its own git checkout of this repo, created directly via the /schedule skill rather than dispatcher.py.
TechEnglishVocab/ is its own top-level folder, not a TechResearch/ subfolder — it's a vocabulary list, not tech research, so it's deliberately left out of the weekly synthesis agent's read scope (synthesis.md (unresolved)): a flat alphabetical word list has no cross-topic connections worth spending synthesis tokens on.
⚠️ If this folder is ever moved or renamed again, the cloud routine's own prompt/scope-guard (configured separately via /schedule, not stored in this repo) must be updated to match — otherwise it keeps writing to whatever path it was last configured with, out of sync with this repo.
What it does
Every Saturday morning, the routine:
- Searches for 4-8 recent, notable pieces of tech content — arXiv papers (cs.AI, cs.CL, cs.SE, cs.LG, cs.DC, etc.), technical whitepapers, or reputable tech-innovation news, prioritizing software development, AI/ML engineering, and research. Hard recency rule: source must be published/submitted within the last 2 years of the run date — verified against the actual publish date, not assumed from search ranking. Anything older is skipped, no matter how notable.
- Actually reads and understands them (not just titles), keeping the exact source URL for each one it draws vocabulary from.
- Extracts 5-15 genuinely tech-context vocabulary words worth knowing cold — skips filler words and product/model proper nouns.
- Appends new rows to the
## Vocabularytable inTechEnglishVocab/tech_english_vocab.md, checking for existing entries first (skip true duplicates, add a new row for a genuinely different tech-context sense of an existing word). Each new word is linked to its source in the first column:[Word](source-url) (type). The meaning/example column is written as**Meaning**: ...<br>**Examples**: ...— stacked on separate lines within the cell (viewer/lib/markdown.tsenablesallowDangerousHtmlspecifically so this<br>renders correctly in the wiki view). - Re-sorts the whole table alphabetically by word.
- Logs the run: prepends a
## [YYYY-MM-DD] Weekend Vocab Scraperblock toLogs/2026.mdwith a[VOCAB]bullet — either the words added, orno qualifying words this runon a quiet week. This happens every run, unconditionally. - Commits — the vocab table (if it changed) and
Logs/2026.md(always) together — and pushes directly to master — no PR gate, matching how Hermes' own nightly job already commits directly to this repo. Commit message isvocab: weekly tech vocabulary update (N words)when words were added, orvocab: log entry, no new words this runon a quiet week.
There's always a commit, even on a quiet week — the log entry is real progress even when the vocab table doesn't change. See Dashboard's "Scheduler Activity" section for a running count of [VOCAB] log entries.
Configuration
| Field | Value |
|---|---|
| Routine name | Weekend Tech Vocab Scraper |
| Routine ID | trig_01P5Jo5FFHfRk8kPoiSdwyFi |
| Cadence | 0 6 * * 6 — every Saturday, 06:00 UTC, permanently fixed (reads as 2:00am EDT right now; will read as 1:00am once winter EST kicks in — not re-adjusted for DST). |
| Model | claude-sonnet-5 |
| Repo | https://github.com/fullstackfusions/PKM (master branch, direct push) |
| Environment | Anthropic cloud default (env_01S8gzx6qMrwgKjVjq6ygeiD) |
| Tools | Bash, Read, Write, Edit, Glob, Grep, WebSearch, WebFetch |
| Scope guard | Only ever modifies TechEnglishVocab/tech_english_vocab.md |
| Manage / view runs | https://claude.ai/code/routines/trig_01P5Jo5FFHfRk8kPoiSdwyFi |
To change the day/time, model, or prompt: edit via the /schedule skill (list → update), or the routines page above. Routines can't be deleted via the API — only through https://claude.ai/code/routines.
Note: since this runs in an isolated cloud sandbox, it has no access to the local machine — it operates purely on its own clone of the GitHub repo, so its changes only reach this vault once pushed and pulled locally (or picked up by whatever already syncs master here, e.g. viewer/deploy/rebuild.sh's pull-on-rebuild).
Related
- tech_english_vocab — the document this routine maintains
- AgentStack/Agents/hermes-harness-spec — the separate, local nightly/weekly cron pipeline (not used for this task)