Competitor diff watch

Monitors competitor websites, changelogs, pricing pages, and sitemaps for changes and produces a weekly diff digest — new pages, messaging shifts, feature launches, pricing moves — with before/after evidence. Use when the user mentions competitor monitoring, competitive intel, "what did [competitor] just launch," pricing change alerts, or tracking competitor content strategy.

Your competitors publish their strategy in public, one page at a time — new sitemap sections before campaigns, pricing-page tweaks before the sales team hears about them, changelog entries the day a feature ships. Almost nobody reads it systematically. This skill reads it weekly and reports only the deltas, because the delta is the intel.

When to run

Recurring: weekly sweep across the configured competitor set (recommended)
On demand: "what changed on Vantiq's site this month" / before competitive strategy reviews

Computer scopes

Config doc: the competitor watchlist (domains, changelog URLs, pricing pages, sitemap indexes)
Snapshot store: this skill's prior crawls (the diff baseline, kept in Computer)
CRM: competitive-deal tags, to correlate moves with pipeline where present

Workflow

Fetch the current state of each watched surface: sitemap indexes (URL inventory and new patterns), pricing pages, changelogs, and key marketing pages. Respect robots directives; public pages only, polite crawl rates.
Diff against the stored baseline: new URLs (cluster by path pattern — a new /agents/ section is a strategy signal, one blog post is Tuesday), removed pages, and text-level changes on watched pages with before/after excerpts.
Classify each delta: product launch, pricing move, messaging shift, content-strategy pattern (new template cluster), or routine. Routine gets counted, not narrated.
Correlate significant moves with competitive-deal activity in the CRM where tagged.
Write the digest: leads with the significant moves and what they plausibly signal — framed as inference with evidence, never as certainty. Store the new baseline.

Guardrails

Autonomy: L0. Observes public surfaces; never interacts with competitor products, never misrepresents identity, never touches gated or authenticated content.
Interpretations are labeled inference; before/after excerpts are the evidence of record.
Excerpts stay short and attributed — evidence, not wholesale copying.

Output

Weekly digest: significant-move blocks (what changed, evidence excerpt, plausible signal, related pipeline note) → new-URL pattern summary → routine-change counts → watchlist health.

Eval fixtures

Set: 3 snapshot pairs of sandbox competitor sites with 22 seeded changes (5 significant, 17 routine) including one new template cluster.
Rubric: diff recall on seeded changes; significant/routine classification agreement; pattern-cluster detection; inference labeling present on all interpretations.
Ship gate: diff recall ≥ 0.95; classification agreement ≥ 0.85.

Run competitor diff watch on your own data.

MORE SKILLS

GTM Engineering & Marketing Ops · L0

ABM account brief

Builds a target-account research brief for ABM plays by merging internal history from GBrain (prior touches, closed-lost context, existing relationships) with external signals (news, hiring, tech stack, filings). Use when the user mentions account research, ABM targets, "brief me on this prospect," territory planning, or prepping outbound to a named account.

GTM Engineering & Marketing Ops · L0

ABM account brief

Builds a target-account research brief for ABM plays by merging internal history from GBrain (prior touches, closed-lost context, existing relationships) with external signals (news, hiring, tech stack, filings). Use when the user mentions account research, ABM targets, "brief me on this prospect," territory planning, or prepping outbound to a named account.

GTM Engineering & Marketing Ops · L2

ICP lead scorer

Scores inbound leads and accounts against the written ICP definition with a stated reason code per score — firmographic fit, use-case signal, engagement quality — and routes by band. Use when the user mentions lead scoring, ICP fit, inbound qualification, MQL routing, "is this lead any good," or wants the inbound queue prioritized.

GTM Engineering & Marketing Ops · L2

ICP lead scorer

Scores inbound leads and accounts against the written ICP definition with a stated reason code per score — firmographic fit, use-case signal, engagement quality — and routes by band. Use when the user mentions lead scoring, ICP fit, inbound qualification, MQL routing, "is this lead any good," or wants the inbound queue prioritized.