CAPTURE auto market-intel
What your market said this week, in 200 words
The file
take it and run it.
One markdown file: the goal prompt, the verifier prompt, and the stopping condition. Read it before you run it. MIT licensed, no account, nothing to sign up for.
curl -fsSL https://vibemarketers.ai/loops/market-intel.md -o .claude/loops/market-intel.mdThe loop
copy it.
Competitors reposition quietly. The homepage changes, the pricing page moves, and nobody on your side notices for a quarter because nobody is looking on a schedule.
Use it when you have three to ten competitors worth watching and no process for watching them, which is most teams.
You need a watchlist of URLs. A browser is enough; there is no tool to buy.
You get a short dated report of what changed since the last run, with every observation naming its source and the day it was read. Unchanged sources are reported as unchanged rather than omitted.
The gate is nothing, because it publishes nothing. The output is internal and the loop is not allowed to assert a competitor result, only what the competitor published.
Goal prompt
Read the watchlist for <company>. For each source, record what it says now
and compare against what it said at the last run. Report only what changed,
with the date each source was read. Write no more than <limit> words. Every
observation names its source and its collection date. Never report a
competitor's results, only what they publish about themselves.Verifier prompt
Read the report. FAIL if it exceeds the word limit, if any observation has
no source and collection date, if it asserts a competitor result rather than
what the competitor published, if it uses causal language, or if it repeats
last week without saying so. Otherwise PASS.Stopping condition
Every source on the watchlist has been read and dated, the report is within the word limit, and the verifier marks PASS.
MCPs it needs
- none required, a browser is enough
Want everything at once
The installer writes the three starter skills and every
published loop into .claude/. Read the
script first.
curl -fsSL https://vibemarketers.ai/install.sh | bashThe receipt
- Baseline
- A watchlist in the Brain naming two format references, both last read on 2026-09-18, with no record of what either said on that date beyond a prose note.
- Hypothesis
- Re-reading a small watchlist on a schedule will catch positioning changes that a person scanning casually would miss, because the comparison is against a recorded state rather than a memory.
- Intervention
- Re-read both sources with a browser and compared what each says now against what was recorded on 2026-09-18.
- Result
- 2 sources read and 2 changes recorded. One competitor now titles itself a private team skills marketplace, where ten days earlier it led with building and sharing skills across a team. The other lists 50 or more loops with engineering-weighted categories and one marketing bucket.
- Commercial signal
- none
- Attribution limits
- This records what a site says about itself, not whether any of it works. Nothing here is evidence about traffic, adoption or revenue at either company, and a homepage title change may reflect a test rather than a decision.
- Learning
- The useful signal was a single word. Private reframes a public marketplace into internal governance, which moves that product away from this one rather than toward it. A summary that only captured topics would have missed it.
- Next intervention
- Record the full homepage text each run rather than a summary, so the diff is mechanical and a one word change cannot be missed by whoever writes the note.
Sources: run log, lab/runs/2026-09-28-market-intel.json, collected , window: 2 sources, compared against readings from 2026-09-18 .
One entry per run, append-only, newest first. Corrections append a dated line.
Inside a company
where the lane changes.
VibeMarketers tests the loop. SuperMarketers runs it on a real team.
- Problem
- Nobody has time to check what competitors published, so positioning drifts against a market nobody is watching.
- Source of truth
- A named watchlist, and the previous run so the report can say what changed.
- Permitted actions
- Read public pages, record what they say, report changes. Never assert a competitor result, never guess at private numbers.
- Required approvals
- Nobody. This reports rather than publishes, and its output is internal.
- Owner
- Marketing lead
- Metric
- Positioning changes caught within one week of the competitor making them.
On a real team the watchlist includes pricing pages and changelogs, where positioning shifts show up before the homepage catches up.
Running this on your own site is one person reading a diff. Running it inside a company changes four things, and the guide covers them. When it stops being a file you copy and becomes a build, SuperMarketers does the same loop model inside your stack, on your keys.