weekly-digest

Weekly Digest

Produce a weekly synthesis from the archived brief notes. All data comes from .hermit/briefs/ — no re-fetching of sources. Cheap to run.

Steps

1. Read the week's briefs. Glob .hermit/briefs/*.md (exclude the weekly/ subdirectory), then filter to files whose filename date prefix falls within the past 7 days (today minus 6 days through today). Read each matching file (frontmatter + body).

   • Directory absent or empty → deliver "No briefs archived this week." and stop.
   • Fewer than 3 briefs → note the gap, produce the digest from whatever is available.
   • Track the actual count of files read — this is brief_count for the archive.

2. Synthesize content. From the brief bodies:

   • Top stories — items appearing across multiple briefs or escalating in coverage. Cite which day(s) they appeared.
   • Emerging themes — topics absent early in the week that dominated late in the week.
   • Faded stories — topics in ≥2 consecutive early-week briefs that appear in no late-week brief. Surface at most 2; omit the section if none meet the threshold.
   • Category activity — count items per category directly from brief bodies across all briefs. Do NOT use top_categories frontmatter for this tally — it is pre-truncated to top 5 per brief and undercounts categories that consistently rank below the cutoff.

3. Source performance. Read feed-sources.md for the full source list, then from frontmatter across the week:

   • Reliable sources — appeared in sources_used in ≥5 briefs.
   • Unreliable sources — appeared in sources_skipped more often than sources_used. Include the skip count (e.g. "Source X (skipped 8/12 briefs)").
   • Silent sources — in feed-sources.md but in neither sources_used nor sources_skipped across any brief this week (may never have been attempted). Keep to 3–5 bullets, actionable framing.
   • sources_skipped (fetch failures) is the reliability signal. sources_quiet (0 items on a quiet day) is NOT — never count quiet days against a source's reliability.