ignitionbot-trend-radar-ops

name	ignitionbot-trend-radar-ops
description	Run IgnitionBot trend sweeps and return scored briefs across configured sources.
memory_tags	["domain:trend-research","workflow:morning-radar","skill_role:researcher","repo_boundary:tools","source:multi-source","outputs:scored-brief","risk:medium"]

IgnitionBot Trend Radar Ops

Use this skill when the task should run the IgnitionBot CLI rather than manually checking each source. IgnitionBot is expected to be either installed on PATH or available as a source checkout pointed to by IGNITIONBOT_DIR.

Inputs

• Query or query list, such as AI agents, developer tools, or creator workflow automation.
• Optional source list: youtube, github, producthunt, hackernews, reddit, polymarket, bluesky, digg, v2ex, rss, web, tiktok, instagram, x, threads, pinterest, linkedin, truthsocial, exa, perplexity, jina, parallel, bilibili, xiaohongshu, douyin, wechat, weibo, xueqiu, xiaoyuzhou.
• Optional window: 24h, 7d, 30d.
• Optional output format: markdown or json.
• Optional watchlist management request.

Workflow

1. Define the candidate universe:

   • Record the query list, window, requested sources, fetch time, and any per-source limits.
   • For open-ended sweeps, include at least one canonical discovery surface appropriate to the lane (for example GitHub, Product Hunt, Hacker News, RSS, or web) rather than relying only on personalized/watchlist sources.
   • Treat the configured source/query matrix as the declared universe for this run; a watchlist alone is not a claim about the broader market.

2. Check availability:

python3 skills/ignitionbot-trend-radar-ops/scripts/run_ignitionbot.py doctor

3. Run a sweep. Prefer explicit queries when the user gives them:

python3 skills/ignitionbot-trend-radar-ops/scripts/run_ignitionbot.py sweep \
  --query "AI agents" \
  --window 30d \
  --sources github,hackernews,reddit \
  --format markdown

4. For recurring topics, use the watchlist:

python3 skills/ignitionbot-trend-radar-ops/scripts/run_ignitionbot.py watchlist add "AI agents"
python3 skills/ignitionbot-trend-radar-ops/scripts/run_ignitionbot.py watchlist list

5. Resolve canonical primary evidence for each shortlisted cluster:

   • Follow secondary reactions back to the official repository, launch, announcement, or direct discussion page.
   • Merge aliases and duplicate results across sources before comparing scores.

6. Run the coverage gate before summarizing:

   • Compare requested sources with sources actually queried and successful.
   • Audit every cluster above the chosen ignition/traction threshold as included or intentionally omitted.
   • Record omitted clusters with a reason and report unavailable, credential-gated, rate-limited, or empty sources as source gaps rather than zero signal.

7. Summarize the brief by leading with:

   • top clusters by ignition score,
   • source evidence,
   • unavailable sources or credential gaps,
   • next action for each high-signal cluster.

Output Contract

Return:

• The command run.
• Top scored clusters with traction score, ignition score, confidence, and action.
• Evidence links from the generated brief.
• Missing credential/source notes when present.
• The declared candidate universe and a compact omitted-candidate audit for high-signal clusters.
• Suggested follow-up sweep only when it directly improves the user's stated research goal.

Guardrails

• Do not hardcode local paths. Use IGNITIONBOT_DIR or an installed ignitionbot executable.
• Do not paste private database paths, credentials, or local config values into the response.
• If optional sources are missing credentials, report that as source coverage, not as a failure.
• X reads must use the Bird-style AUTH_TOKEN + CT0 session GraphQL adapter. Do not use an X developer bearer token or fall back to the paid official recent-search API. Report a stale session or query ID as a source gap.
• Do not promote a secondary reaction without resolving the canonical event when a primary source is available.
• Keep the answer focused on trend evidence and actions, not raw JSON dumps, unless the user explicitly asks for JSON.