gtmskills.com

Reference file

Triage

triage.md

Triage rubric — urgency, action badges, mischaracterization records

Apply to every signal, every run. A signal without all four labels (urgency, action, type, exact URL) is not done being triaged.

Urgency levels

  • 🔴 Act now — a mischaracterization going viral, high-reach negative coverage, a competitor counter-announcement, or a crisis-shaped signal. The team should see it today.
  • 🟡 Monitor — an emerging negative theme, mid-reach inaccurate coverage, competitor positioning content. Track for escalation; no action yet.
  • 🟢 Good signal — accurate positive coverage, organic enthusiasm, adoption signals. Candidates for amplification.
  • ⬜ Noise — irrelevant mentions, same-name products, spam. Filtered by the materiality threshold below; never shown in the main feed.

Urgency is a function of reach × wrongness × velocity, not tone alone. A scathing post with 12 views is Monitor at most; a politely wrong claim in a high-reach outlet is Act now.

Reach heuristics (when platforms don't publish numbers)

  • Hacker News: 50+ comments = high priority; front page = treat as press-tier reach
  • X: 100+ likes within 24h = meaningful; quote-posted by a journalist or known account = escalate a level
  • Reddit: top-of-subreddit in a large sub, or 100+ upvotes = meaningful; comment depth signals real engagement over drive-by voting
  • YouTube/TikTok: view velocity in the first 48h matters more than the absolute count
  • Press: use the outlet's typical audience as a proxy; secondary pickup (another outlet citing the piece) is worth more than raw readership
  • State reach as an estimate ("~50K readers", "front page of HN") and mark it unverified when you could not confirm it

Action badges

  • RESPOND — needs a direct public response: a reply in the thread, a comment, press outreach
  • CORRECT — needs a correction or clarification: a DM, a comment, a note to the journalist
  • AMPLIFY — worth sharing, reposting, or building on
  • ESCALATE — belongs with comms, legal, or leadership — not with whoever is watching the feed
  • WATCH — no action yet; track for escalation and name what would trigger it
  • IGNORE — filtered noise (record it in the appendix, not the feed)

Every RESPOND/CORRECT badge must come with a specific suggested action — the actual sentence to post and where to post it. "Consider responding" is not an action. All suggested actions are drafts for a human to approve and send.

Signal types

Press coverage · Community discussion · Social mention · Competitor move · Mischaracterization · Churn signal · Influencer take · Developer reaction · Analyst comment

Source URL — required for every signal

Every signal carries the exact URL of the specific article, thread, or post, exactly as the search tool returned it.

Correct:

  • https://techcrunch.com/2026/06/11/example-product-launch
  • https://news.ycombinator.com/item?id=12345678
  • https://reddit.com/r/MachineLearning/comments/abc123/thread_title
  • https://x.com/[username]/status/[POST_ID]

Wrong — never use: https://techcrunch.com, https://news.ycombinator.com, https://reddit.com, https://x.com. A homepage link is useless to the reader, and a fabricated deep link is worse than no link — if you don't have the exact URL, say so on the card.

Mischaracterization record — six fields, all required

For every piece of coverage that gets something wrong:

  1. The claim — exactly what was said, quoted
  2. The source — outlet, author, reach estimate, date
  3. What's wrong — the specific inaccuracy, not a vibe
  4. The correct version — the accurate statement, checkable against the announcement
  5. Spread risk — search for secondary coverage citing the wrong claim before assigning this; "is being picked up" requires found evidence
  6. Suggested response — a one-sentence, copy-ready correction

Status badges

  • SPREADING — found in 2+ independent sources, or being cited/quoted onward
  • CONTAINED — single source, no secondary pickup found
  • CORRECTED — the source amended it, or the correction is visibly winning in the thread

Status drives urgency more than the original outlet's size: a small blog's wrong claim that three other places are citing outranks a big outlet's wrong claim nobody repeated. A mischaracterization stays on the tracker across re-runs until its status is CORRECTED — carry it forward even when no new coverage mentions it.

Materiality threshold

Skip entirely (don't triage, don't count):

  • Clearly a different product with the same name (the context profile is the filter)
  • Automated bots or spam accounts
  • Duplicate coverage with no new information — syndicated wire copy counts once, at the original
  • Single-digit reach with no amplification potential

Flag but don't prioritize:

  • Neutral coverage that accurately describes the product
  • Positive signals that need no action