agentleFS
duncan-buildroomduncan-buildroom/profit-room-skills/skills/daily-digest/SKILL.md

daily-digest

Skill18 starsChanged 4 months agoMITView on GitHub

Run a daily industry digest for the user's niche — aggregates fresh tool launches, releases, and high-signal discussions from public sources (Reddit, Hacker News, GitHub, web search), filters hard, and outputs a tight briefing. No fluff, no opinion pieces, just specific things the user could act on or make content about today. Use when the user says /daily-digest, 'what's new today', 'what happened in [my niche]', 'morning briefing', 'catch me up', or 'what's trending'.

17 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

9 rules. 0 are common, 4 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 11 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If it doesn't exist, run the personalize skill first (or just tell me your niche and who you serve).

    line 18
  2. Very specific:

    No flag → DAYS=2 (last 48 hours)

    line 41
  3. Distinctive:

    Use timeout 300000ms (5 minutes).

    line 136
  4. Very specific:

    DEDUP: Read ~/.claude/digest-seen.txt (lines YYYY-MM-DD|title-slug).

    line 157
  5. Very specific:

    Wave check: if 3+ independent sources reference the same tool within the window, prefix it 🌊 and surface it first — the content window is …

    line 176
  6. Distinctive:

    One-phrase audience note: every DROPS item gets one line reframing it as leverage for the user's audience.

    line 179
  7. Distinctive:

    Every item needs a source link.

    line 189
  8. Very specific:

    Recency is everything: Always use after:{CUTOFF_ISO} in web searches.

    line 253
  9. Very specific:

    If the user's niche isn't software, drop GITHUB_ORGS and lean on Reddit + web search + Product Hunt.

    line 257

“Common” is measured: kept by at least 6 in 10 of the 11 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Daily Digest — Industry Briefing

Surfaces specific, concrete launches, releases, and discussions in the user's field that they could make a single piece of content about, or just need to know. No corporate news, no debates, no "AI is changing everything" filler.

Personalization

This digest is scoped to your niche and your audience. Before running, load the brand profile at ~/.claude/brand-profile.md.

  • If it doesn't exist, run the personalize skill first (or just tell me your niche and who you serve).
  • NICHE and AUDIENCE come from the profile and drive every search query and the relevance filter.
  • The digest is tuned for a solo creator/operator by default — someone trying to grow an audience, get clients, or ship products. If the user's audience is different, adapt the "relevance" test to their audience's core question.

Configuration (edit these for your niche)

Set these once for your field. The defaults below assume an "AI / creator tools" niche — swap them for your own. You can paste your own values when you run the skill, or keep them here.

  • SUBREDDITS — communities where your niche talks (default: ClaudeAI ChatGPT SideProject indiehackers entrepreneur)
  • GITHUB_ORGS — orgs whose new repos = signal, if your niche is software (default: anthropics openai google-deepmind; leave empty for non-software niches)
  • KEYWORDS — the tools/terms a launch would mention (default: AI agent LLM open source tool)
  • COMPETITOR_CHANNELS — optional: YouTube channel IDs of creators you track (used only if vidIQ is connected; empty by default)

STEP 0: Parse Timeframe

  • --today / --days=1 → DAYS=1
  • --week / --days=7 → DAYS=7
  • --days=N → DAYS=N
  • No flag → DAYS=2 (last 48 hours)

Compute:

  • TODAY_ISO — today as YYYY-MM-DD (from context)
  • CUTOFF_ISO — TODAY_ISO minus DAYS days

Display before running tools:

🗞️ Running your [NICHE] digest for the last {DAYS} day(s)...
Scanning: Reddit · Hacker News · GitHub · web search

STEP 1: Run Research In Parallel

Fire the Bash call and the WebSearches in ONE message. Replace {DAYS}, {CUTOFF_ISO}, and the config variables with real values before firing.

Tool call A — one Bash call across the free public APIs:

DAYS={DAYS}
CUTOFF_ISO={CUTOFF_ISO}
SUBREDDITS="ClaudeAI ChatGPT SideProject indiehackers entrepreneur"   # ← your SUBREDDITS
GITHUB_ORGS="anthropics openai google-deepmind"                        # ← your GITHUB_ORGS (may be empty)
KEYWORDS="AI agent LLM open source tool"                               # ← your KEYWORDS
CUTOFF_EPOCH=$(python3 -c "import datetime; print(int((datetime.datetime.now()-datetime.timedelta(days=$DAYS)).timestamp()))")

# ── Reddit hot posts (real upvote signal; score >= 150 is strong) ──
echo "=== Reddit Hot Posts ==="
for sub in $SUBREDDITS; do
  echo "--- r/$sub ---"
  curl -fsSL -H "User-Agent: daily-digest/1.0" \
    "https://www.reddit.com/r/${sub}/hot.json?limit=10" 2>/dev/null \
    | python3 -c "
import sys, json
try:
    data = json.load(sys.stdin)
    for post in data['data']['children']:
        p = post['data']
        if p.get('score', 0) >= 150:
            print(f\"{p['score']}up  {p['title'][:120]}  |  https://reddit.com{p['permalink']}\")
except Exception as e:
    print(f'WARN: {e}')
" 2>/dev/null || echo "WARN: r/$sub fetch failed"
done

# ── Hacker News (Show HN + discussions, real points, date-filtered) ──
echo "=== HN Show HN ==="
curl -fsSL "https://hn.algolia.com/api/v1/search?query=$(echo $KEYWORDS | tr ' ' '+')&tags=show_hn&numericFilters=created_at_i%3E${CUTOFF_EPOCH},points%3E50&hitsPerPage=15" 2>/dev/null \
  | python3 -c "
import sys, json
try:
    data = json.load(sys.stdin)
    for hit in sorted(data.get('hits', []), key=lambda h: h.get('points', 0), reverse=True):
        print(f\"{hit.get('points',0)}pts | {hit.get('title','')} | https://news.ycombinator.com/item?id={hit.get('objectID','')}\")
except Exception as e:
    print(f'WARN: {e}')
" 2>/dev/null || echo "WARN: HN Show HN failed"

echo "=== HN Discussions ==="
curl -fsSL "https://hn.algolia.com/api/v1/search?query=$(echo $KEYWORDS | tr ' ' '+')&numericFilters=created_at_i%3E${CUTOFF_EPOCH},points%3E100&hitsPerPage=5" 2>/dev/null \
  | python3 -c "
import sys, json
try:
    data = json.load(sys.stdin)
    for hit in data.get('hits', []):
        print(f\"{hit.get('points',0)}pts | {hit.get('title','')} | https://news.ycombinator.com/item?id={hit.get('objectID','')}\")
except Exception as e:
    print(f'WARN: {e}')
" 2>/dev/null || echo "WARN: HN discussions failed"

# ── New GitHub repos from key orgs (software niches only; catches drops at source) ──
if [ -n "$GITHUB_ORGS" ]; then
  echo "=== New GitHub repos from key orgs ==="
  for org in $GITHUB_ORGS; do
    echo "--- $org ---"
    curl -fsSL "https://api.github.com/orgs/${org}/repos?sort=created&direction=desc&per_page=10" 2>/dev/null \
      | python3 -c "
import sys, json
org='${org}'; cutoff='${CUTOFF_ISO}'
try:
    for r in json.load(sys.stdin):
        created = r.get('created_at','')[:10]
        if created >= cutoff:
            desc = (r.get('description') or '')[:80]
            print(f\"{created} | {org}/{r['name']} | ★{r.get('stargazers_count',0)} | {desc} | {r['html_url']}\")
except Exception as e:
    print(f'WARN: {e}')
" 2>/dev/null || echo "WARN: $org fetch failed"
  done
fi

Use timeout 300000ms (5 minutes).

Tool calls B–F — WebSearches (fire in the same message). Replace [NICHE]/[KEYWORDS] with the user's values and always anchor recency with after:{CUTOFF_ISO} (a bare "this month" returns the same cached results every day):

  1. [NICHE] new tool OR feature OR release OR update after:{CUTOFF_ISO}
  2. [NICHE] "just launched" OR "just shipped" OR "just released" after:{CUTOFF_ISO} -enterprise
  3. [main tool/platform in the niche] new feature OR changelog after:{CUTOFF_ISO}
  4. site:producthunt.com [NICHE] launch after:{CUTOFF_ISO}
  5. Discovery (no keyword lock): "just released" OR "open source" [broad niche term] after:{CUTOFF_ISO} — catches tools before you know their names

Optional — competitor videos (only if vidIQ MCP is connected and COMPETITOR_CHANNELS is set): Load the schemas first with ToolSearch (select:mcp__claude_ai_vidIQ_for_Claude__vidiq_channel_videos,mcp__claude_ai_vidIQ_for_Claude__vidiq_outliers,mcp__claude_ai_vidIQ_for_Claude__vidiq_video_comments), then pull recent uploads (popular=false, 7-day window) and outliers for each channel, and scan the comments on any tool-specific video for questions/pain points. If vidIQ isn't available, skip the COMPETITORS and TOP COMMENTS sections entirely.


STEP 2: Filter Hard

DEDUP: Read ~/.claude/digest-seen.txt (lines YYYY-MM-DD|title-slug). Skip any item whose title (lowercased, spaces→dashes) or URL matches an entry from the last 7 days. If the file doesn't exist, continue without dedup.

EXCLUDE anything that is: safety/alignment/regulation debate, corporate earnings/funding/valuations, enterprise B2B that doesn't touch a solo creator, geopolitics/policy, general "X is changing everything" opinion, or any item whose only angle is "put this on your resume / land a better job." (The default audience is trying to build something of their own, not climb a ladder.)

INCLUDE only: a specific new tool/feature/release, something a solo creator could try or demo today, a concrete workflow or technique that changes how you work, or a named product launch and what it does.

Relevance test (calibrate to the profile's AUDIENCE): "Can someone in my audience use this today to grow their audience, get clients, or make something?" If yes, it belongs. If it's for a 500-person enterprise, cut it.

First-mover check: for each launch, note if no creator content exists on it yet — mark it 🟢. Those rank higher; the content window is open.

Wave check: if 3+ independent sources reference the same tool within the window, prefix it 🌊 and surface it first — the content window is closing fast.

One-phrase audience note: every DROPS item gets one line reframing it as leverage for the user's audience. One phrase, not a paragraph. If you can't write an honest one that passes the relevance test, the item doesn't belong in DROPS.


STEP 3: Output the Digest

Exactly these sections, in order (omit any that are empty): DROPS, SIGNAL, and — only if vidIQ was used — COMPETITORS and TOP COMMENTS. Quality over quantity: 2 sharp items beat 5 vague ones. Every item needs a source link.

🗞️ [NICHE] DIGEST — {Day}, {Date} · {DAYS}d

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔥 DROPS

→ ⭐ 🆕 *{Tool}* — {what it is / changed, one phrase}  [source](URL)
  ↳ {one-phrase audience note}
→ 🟢 *{Tool}* — {…}  [source](URL)
  ↳ {audience note}

Labels: ⭐ lead item (the one thing to act on today, exactly 1) · 🟢 first mover · 🌊 wave forming · 🆕 new release
[3-5 items max. Omit the section if nothing genuinely dropped.]

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⚡ SIGNAL

→ r/{sub} {score}↑: "{post title}"  [reddit](URL)
→ {N}pts HN: "{Show HN title}"  [hn](URL)

[Reddit hot threads + HN discussions. Max 7. Show scores. Skip anything without something specific.]

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🔭 COMPETITORS   (only if vidIQ used)

→ {Channel}: "{Video title}" — {views}/{days}d ↑outlier  [yt](URL)

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
💬 TOP COMMENTS   (only if vidIQ used)

→ "{Verbatim comment}" — @handle on {Channel}
  *Angle: {one video the user could make that answers this}*

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📊 {N} sources · {N} deduped

Pick the single strongest item (first-mover window open + clean relevance + high signal) and mark it ⭐ in DROPS.


STEP 4: Save Dedup State

Append one line per DROPS and SIGNAL item to ~/.claude/digest-seen.txt:

{TODAY_ISO}|{title-slug}

title-slug = title/handle lowercased, spaces→dashes, truncated at 60 chars. Then prune lines older than 7 days.


STEP 5: Follow-Up

---
Go deeper on anything: name it and I'll run a full dive, draft a post, or turn it into a content idea.

Notes

  • Recency is everything. Always use after:{CUTOFF_ISO} in web searches. A bare month/year returns the same cached popular articles regardless of date.
  • Source priority (earlier = more first-to-know): GitHub org watcher → HN Show HN (12-24h before Twitter/Reddit) → web search launches → Reddit hot threads.
  • If the user's niche isn't software, drop GITHUB_ORGS and lean on Reddit + web search + Product Hunt.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.