competitive-intel-watch skill

Competitive Intel Watch

Purpose

Monitor a competitive landscape for material shifts since the last run. Diff the world against the previous snapshot; report only what changed, with evidence; flag which downstream artifacts need updating. This is the skill that turns competitive research from a document into a cadence — the weekly SIGINT sweep and monthly OSINT digest from the fusion cadence live here. A watch reports change, not state: regenerating the same report weekly is theater, and "no material shifts this cycle" is a valid, useful result.

Input

Works best with: the previous Competitive Research Snapshot (pasted or attached) — the baseline this run diffs against — and the competitor list (defaults to those in the snapshot). Also useful: anything specific you're watching for this cycle, and a materiality bar adjustment if the default needs tightening or loosening.

Input supplied inline with the invocation — text after the skill name, a pasted context dump, or an appended ARGUMENTS: line — counts as answers already given. Use it against the question budget; don't re-ask.

Arriving empty-handed? That works too. With no prior snapshot, the skill falls back to baseline mode: it produces a first snapshot using the competitive-research-snapshot structure and stops — the delta value starts on run two.

Example invocation: Competitive intel watch — prior snapshot pasted below; this cycle I'm specifically watching for pricing moves. [snapshot]

Key Concepts

contract — question budget of 2 (this skill's tightest), search-plan gate, Fact/Inference/Assumption labels, Just Enough Mode, stable schema, 4-option Final Step.

• Governing protocol: honors the autonomous-investigation

OSINT and HUMINT signals monthly and FININT on the quarterly pass — the fusion cadence in intelligence-collection-disciplines is this skill's operating rhythm.

• Discipline mix: SIGINT first (site diffs, pricing pages, job posts — the freshest layer), with

plausibly act on it: pricing/packaging changes, launches and deprecations, positioning shifts, leadership moves, funding or M&A, major customer wins/losses, credible roadmap signals. Below the bar: cosmetic site changes, routine content marketing, minor releases. Why it matters: a watch that cries wolf gets ignored by cycle three — the bar is what keeps the audience.

• The materiality bar. Report a change only if a sales rep, pricing owner, or roadmap owner would

regenerate it. The empty changelog is a first-class outcome.

• Delta discipline. Read the prior snapshot fully before searching; diff against it, never

material shift maps to the battle card, positioning, pricing, or roadmap sections now stale.

• Update flags close the loop. Research is only done when it names the artifact it changes — each

product claims. Every claimed change carries a URL and a date.

• Do-not-invent list: competitors, features, pricing, market share, customer wins, roadmap items,

competitive-research-snapshot first; the scope itself changed (new segment, pivot) → re-snapshot from scratch rather than diffing a stale scope.

• When NOT to use: no baseline exists and you want the full treatment → run

Application

snapshot per the competitive-research-snapshot schema and stop.

1. Determine mode. Prior snapshot provided → delta mode. None → baseline mode: produce a

If unanswered, proceed: baseline mode if no snapshot, default materiality bar otherwise.

1. Credit inline context, then ask only the unanswered questions (max 2):
2. Do you have the previous snapshot, or should I create a baseline?
3. Anything specific you're watching for this cycle?

memory of the market.

1. Read the prior snapshot fully before searching. The diff target is the document, not your

pricing pages, release notes, press, investor materials, credible news, review sites, job postings), how facts will be separated from inference. Continue unless revised.

1. Show the 3-bullet search plan — what you'll check per competitor, source types (company sites,

1. Sweep and filter through the materiality bar. When nothing clears it, say so plainly.
2. Emit the schema below exactly — runs must be diffable.

Output schema (do not reorder)

~~~markdown

Competitive Watch Report

1. Run Header

Scope (from prior snapshot): | Prior snapshot date: | This run date: | Competitors checked:

2. Changelog (Material Shifts Only)

For each material shift:

[Competitor] — [4 to 8 word change summary]

• What changed: [1-2 bullets, labeled Fact/Inference]
• Evidence: [URL, date]
• So what: [why it clears the materiality bar]
• Confidence: [high / medium / low]

If nothing cleared the bar: "No material shifts this cycle." List anything on the watchlist for next run.

3. Update Flags

Only rows with real updates; omit the rest.

4. Watchlist for Next Run

• [Signals below the bar but trending]
• [Open questions this run could not resolve]

Assumptions to Validate

~~~

• [Assumption 1] / [Assumption 2] / [Assumption 3]

A copy/paste fill-in version of this schema, with quality checks, lives in template.md.

Final Step (offer exactly 4 options)

1. Update the battle card sections flagged above (battle-card-builder)
2. Deep-dive the most significant change
3. Produce the refreshed full snapshot (new baseline)
4. Adjust the materiality bar or competitor list for next run

Accept 1, 2, 3, 4, 1 and 2, Verbose Mode, or a custom path. On a scheduled, unattended run, file the report and stop — the options wait for a human.

Examples

A changelog entry that clears the bar (fictional):

### Ledgerline — mid-tier plan removed from pricing page

- What changed: the $49 "Team" tier no longer appears; feature list redistributed upward —

Fact (pricing page vs. archived version, Jul 2 vs. Jun 1)

- Evidence: URL + archive diff, dated