| title | Daily Bible Reading Cron |
| name | daily-bible-reading |
| version | 1.0.0 |
| description | Daily cron job that reads one book of the Bible, extracts 3 lessons with practical application to server operations, and appends to SOUL.md. Runs autonomously at 1am — no user interaction, no notification.
|
| triggers | ["Cron job at 1am daily","Reading one book of the Bible per day through the full canon","SOUL.md maintenance and state file management"] |
Daily Bible Reading Cron
Implementation Note
There are two implementations of the daily bible reading cron on this system:
| Aspect | Legacy (this skill) | Active (installed) |
|---|
| Type | LLM-driven cron | no_agent Python script |
| Script | N/A (LLM prompts directly) | ~/hermes-cortex/src/scripts/agent-daily-bible-reading.py |
| API | bible-api.com (WEB translation) | OpenCode Zen (deepseek-v4-flash) |
| State tracking | ~/.hermes/bible-reading-state.txt | Parses SOUL.md for last book read |
The active implementation is the no_agent Python script installed as
agent-daily-bible-reading — see the agent-daily-bible-reading skill for
its current behavior. This skill documents the original LLM-driven pattern
and the state file it introduced.
Legacy Pattern (LLM-driven cron)
Schedule
Runs at 1am daily, autonomous — no user interaction, no notification.
Flow (per run)
- Read the state file to find the current book/position:
cat ~/.hermes/bible-reading-state.txt
- Fetch the next chapter from bible-api.com:
curl -s "https://bible-api.com/<book>+<chapter>?translation=web"
- Extract 3 lessons with practical application to server operations
(the fleet's frame: reliability, discipline, stewardship).
- Append the entry to SOUL.md under the daily bible entries section:
### <Book> — *"<key verse>"* (<Book> <chapter>:<verse>)
<lesson text with application>
- Advance the state file to the next book/chapter through the full canon.
State file management
cat ~/.hermes/bible-reading-state.txt
echo "Genesis:1:1" > ~/.hermes/bible-reading-state.txt
echo "Exodus:3:1" > ~/.hermes/bible-reading-state.txt
Transition History
The legacy LLM cron was replaced because the small local model produced
unreliable lesson quality every tick. The active no_agent script uses a
single API call to OpenCode Zen (deepseek-v4-flash) for the lesson generation
and handles the rest deterministically — the cron-no-agent-conversion
pattern applied to scripture reading.
Spurious-Run Recovery (double-entry accident, 2026-08-10)
The active no_agent script ignores unknown args — probing it with
--help (or any flag) executes the FULL side-effect path: it appends a new
### Book — entry for the NEXT book, and when the section exceeds the bound
it archives the oldest entries out of SOUL.md. Two probe runs in one session
produced two same-day entries (Deuteronomy + Joshua) and evicted the
legitimate Leviticus + Numbers entries — jumping the cycle anchor two books.
Recovery (verify BEFORE the next scheduled 01:00 run):
- Restore the evicted SOUL.md entries verbatim from
~/brain/<agent>/bible/archive/SOUL-archive.md (the script appends the
blocks it evicts — copy the exact ### Book — block back into
## Scripture Insights)
- Delete the premature brain page
~/brain/<agent>/bible/<book>.md and
revert the INDEX.md row it stamped to —
- Trim the duplicate archive section the spurious run appended
- Verify the LAST
### Book — entry in SOUL.md is the correct anchor
(Numbers → next should be Deuteronomy, not Joshua)
Do NOT re-fire the cron via the scheduler to clear a last_status: error
that came from a transient API timeout when the day's entry was already
recovered — re-firing stamps the NEXT book with today's date, shifting the
whole cycle. Let the natural next scheduled run clear the status.
Related
agent-daily-bible-reading — the active installed implementationcron-no-agent-conversion — why/how the conversion happenedsoul-refinement — SOUL.md maintenancecron-job-management — cron naming and lifecycle
