content-digests

name	content-digests
description	Build and deliver periodic content digests (news, stock, entertainment) to multiple platforms: save to vault Raw/ and print full markdown to stdout for cron-based delivery.
version	1.0.0
author	Hermes Agent
license	MIT
metadata	{"hermes":{"tags":["content","digests","news","aggregation","multi-platform"]}}

Content Digest Pipelines

This umbrella skill governs the construction and delivery of periodic content digests — news, stock updates, entertainment listings, and similar recurring data aggregations — to multiple platforms (Discord, Telegram, Slack) while archiving to an Obsidian vault.

Core Principles

1. Dual output — Every digest script must:

   • Save full markdown to the vault's Raw/<Category>/ directory for permanent archival
   • Print the complete markdown to stdout for cron-based platform delivery

2. Platform-agnostic — The same script drives all delivery channels. Cron configuration decides which platforms receive the stdout capture. No per-platform script branching.

3. Chunking for size-limited platforms — Telegram (4096 char limit) and Slack (~4000 chars) require splitting large digests into sequential chunks with part numbering. Discord handles larger messages (~8KB) but chunking still improves UX.

4. Vault-first organization — All content lives in Raw/<Category>/<Date> <Digest Name>.md inside the Obsidian vault. The filename convention is YYYY-MM-DD <Digest Name>.md.

When to Use

Load content-digests when:

• Building a recurring news or data aggregation script
• Setting up periodic delivery to Discord/Telegram/Slack
• Ensuring content is both viewable in-channel and persisted to knowledge base
• Refactoring a script that only saves to file but doesn't print content

Trigger Conditions

• User asks for daily/weekly digest of any domain (tech news, stocks, entertainment)
• Existing digest script only writes to file without stdout delivery
• Need to add a new platform channel (Telegram, Slack) to an existing digest
• Fixing "content not appearing in channel" complaints

Implementation Recipe

Step 1 — Vault paths

from pathlib import Path
import os, datetime

VAULT = Path.home() / "obsidian-vault" / "FACorreia"  # adjust vault name
RAW_DIR = VAULT / "Raw" / "<Category>"          # e.g. Raw/Dev/Swift, Raw/Entertainment
RAW_DIR.mkdir(parents=True, exist_ok=True)
now = datetime.datetime.now()
outfile = RAW_DIR / f"{now.strftime('%Y-%m-%d')} <Digest Name>.md"

Step 2 — Build markdown (md string)

Consistent header:

md = f"""# <Digest Name> — {now.strftime('%Y-%m-%d %H:%M')}

> Summary line with counts per source
---
"""
md += "\n".join(entries)  # your collected items
md += f"\n\n✅ Saved to `{outfile}`"

Step 3 — Save to vault (always)

with open(outfile, 'w') as f:
    f.write(md)
print(f"✅ <Digest> saved ({len(entries)} items) → {outfile}")

Step 4 — Print to stdout for cron delivery

Critical: This must be the LAST print statement, outside all loops, with no extra formatting before or after (except optional separator).

# Separator helps platform delivery bots distinguish payload
print("\\n" + "="*60)
print(md)   # FULL markdown content

Pitfall: Double-printing (inside/outside loops) causes duplicate messages. Ensure only one print(md) call after saving.

Step 5 — Chunking (for Telegram/Slack)

If the script is invoked per-platform with size limits, wrap the final delivery in a chunker:

def chunk_send(content, limit=3900):
    """Yield chunks ending at newline boundaries."""
    lines = content.split('\n')
    chunk = []
    size = 0
    for line in lines:
        line_nl = line + '\n'
        if size + len(line_nl) > limit and chunk:
            yield ''.join(chunk)
            chunk = [line_nl]
            size = len(line_nl)
        else:
            chunk.append(line_nl)
            size += len(line_nl)
    if chunk:
        yield ''.join(chunk)

full_payload = "\\n" + "="*60 + "\\n" + md
for idx, part in enumerate(chunk_send(full_payload), 1):
    header = f"🔹 <Digest> (Part {idx}/{})" if total>1 else f"🔹 <Digest>"
    print(header + "\\n\\n" + part)

However, prefer letting the cron delivery bot handle chunking by delivering the full stdout to each configured platform. Only implement manual chunking in the script if the bot doesn't support automatic splitting.

Digest-Specific Conventions

Digest	Vault Category	Channel (Discord)	Frequency	Sources
Swift News	Raw/Dev/Swift/	#swift-news (1499908671847661578)	Daily 09:00	Hacking with Swift, Swift Forums, HN, GitHub
Go News	Raw/Dev/Go/	#go-news (1499908914500862123)	Daily 09:00	Go Blog, Go Forum, HN, GitHub
Stock News	Raw/Dev/Stock/	#stock-plan (1499338003334561843)	Daily 09:00	Yahoo Finance, Motley Fool
Entertainment	Raw/Entertainment/	#entertainment (1499331939469889656)	Daily/Weekly	TVmaze, MyAnimeList, IMDb
IMDb Weekly	Raw/Movies/, Raw/TV/Seasonal/, Raw/Anime/	#entertainment	Weekly (Fri)	IMDb datasets

Validation

# 1. Script runs without error
python3 digest_script.py

# 2. File saved to vault
ls -l /opt/data/obsidian-vault/FACorreia/Raw/Dev/Swift/$(date +%Y-%m-%d)\ *.md

# 3. Stdout contains full markdown (not just summary)
python3 digest_script.py | head -c 500   # should show content, not just "✅ saved"

# 4. Cron job configured with correct model (for LLM-backed digests)
hermes cron list | grep <digest-name>
# Model should be anthropic/claude-sonnet-4.6 (NOT openrouter/anthropic/claude-sonnet-4)

Common Pitfalls

Pitfall	Symptom	Fix
print(md) inside a loop	Duplicate content in channel	Move print(md) to AFTER the loop, at root scope
Forgetting stdout entirely	Cron delivers only the summary line	Add print(md) after file save
Printing debug to stdout	Debug text appears in Telegram/Discord	Send diagnostics to sys.stderr or a log file
Platform 400s due to oversize	Telegram rejects message (>4096 chars)	Chunk output or reduce digest size
Wrong output directory	File saved but cron can't find it	Ensure stdout is captured, not just file I/O
Security restrictions on piped commands	Command fails with security warning	Use intermediate file: script.py > /tmp/output && script2.py < /tmp/output instead of `script.py

Cron Job Template

hermes cron create "0 9 * * *" \
  --name "daily-swift-news" \
  --prompt "Aggregate Swift news and deliver to Discord #swift-news" \
  --script "swift_news_digest.py" \
  --model "anthropic/claude-sonnet-4.6" \
  --deliver "discord:1499908671847661578"

Multi-platform variant — Create separate jobs with same script but different --deliver targets (Discord, Telegram, Slack) to ensure all channels receive the digest:

# Discord
hermes cron create "0 9 * * *" --name "swift-news-discord" \
  --script "swift_news_digest.py" --model "anthropic/claude-sonnet-4.6" \
  --deliver "discord:1499908671847661578"

# Telegram Home
hermes cron create "0 9 * * *" --name "swift-news-telegram" \
  --script "swift_news_digest.py" --model "anthropic/claude-sonnet-4.6" \
  --deliver "telegram:476978568"

# Slack Home
hermes cron create "0 9 * * *" --name "swift-news-slack" \
  --script "swift_news_digest.py" --model "anthropic/claude-sonnet-4.6" \
  --deliver "slack:C0B0BDGEJTT"

Alternatively, a single job with --deliver origin broadcasts to all configured platforms if the gateway is set up to mirror.

References

• references/multi-platform-delivery-pattern.md — stdout-driven delivery and chunking strategies
• references/vault-raw-convention.md — Raw/<Category>/ file organization
• scripts/digest-template.py — minimal working example (see scripts/ directory)