what-changed-in-search
Skill0 starsChanged 4 months agoMITView on GitHub
Check what changed in search or answer engines recently, always fetched live and never answered from training data, then show what it means for this specific site in measured counts with a control check before blaming anything on it.
In this repository26 other files it gives its agents.
- beat-my-competitorsSkill
- brief-my-writerSkill
- can-i-actually-rankSkill
- check-before-i-publishSkill
- explain-this-numberSkill
- find-my-quick-winsSkill
- follow-the-moneySkill
- get-cited-elsewhereSkill
- get-mentioned-by-aiSkill
- know-my-buyerSkill
- make-it-sound-humanSkill
- make-me-quotableSkill
- map-my-siteSkill
- pick-my-battlesSkill
- put-it-on-autopilotSkill
- rainmakerSkill
- revive-old-pagesSkill
- say-it-their-waySkill
- show-me-progressSkill
- show-up-in-communitiesSkill
- spread-one-piece-everywhereSkill
- stop-losing-visitorsSkill
- unblock-my-money-pagesSkill
- what-actually-workedSkill
- what-to-target-nextSkill
- write-the-pageSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 1 is distinctive, and 7 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
7
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 24
Run rainmaker context --check first.
- Very specific: line 32
Only the data/ files listed in this skill's Consumes table.
- Very specific: line 34
If strategy.json.context_hash does not match the current hash of context/business.md, say exactly:
- Very specific: line 40
If context/business.md carries confidence: stub, continue, and stamp every output with: "Built on a stub context.
- Very specific: line 72
Always fetch. Always stamp fetched_at on the finding.
- Distinctive: line 73
Never assert that the update caused a metric change without the control check alongside it.
- Very specific: line 74
Write the algo_update event regardless of whether the site was affected.
- Very specific: line 98
Every claim carries its source tier, the control check is present and answered either way, and an algo_update event was appended to the …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
what-changed-in-search
Context load
Run rainmaker context --check first. It prints what exists, what is stale, and exits 1 if anything this skill requires is missing.
Then read, in this order:
context/business.mdin full. If absent, stop: "No business context. Runrainmaker audit, then theknow-my-buyerskill."context/voice.mdif this skill writes prose. If absent, stop and say so.context/glossary.mdif this skill names products, features or competitors.data/strategy.jsonif this skill reads or writes strategy.- Only the
data/files listed in this skill's Consumes table. Never crawl or call an API the core already covers.
If strategy.json.context_hash does not match the current hash of context/business.md, say exactly:
"Business context was edited after the strategy was written. Re-run know-my-buyer, or run rainmaker context --sync to accept the prose as authoritative."
Then stop.
If context/business.md carries confidence: stub, continue, and stamp every output with: "Built on a stub context. Nothing in it came from a buyer. Run know-my-buyer to replace it."
Consumes
| File | Why | If missing |
| live web | source of the actual update information | this skill never answers from training data; if it cannot fetch, it says so and stops rather than guessing |
data/snapshots/<latest>/crawl.json | which of our pages match the affected pattern | run rainmaker audit |
data/state.json | tiers of the affected pages | run rainmaker audit |
data/snapshots/*/gsc.json (2+ snapshots spanning the update window) | position and impression deltas across the window | run rainmaker fetch --source gsc again if only one snapshot exists |
data/ledger.jsonl | what was shipped in the same window, for the control | run rainmaker ledger --rebuild |
Produces
An algo_update ledger event (site-level, id: "site"), so what-actually-worked can later attribute a decline to a core update rather than to shipped work.
Refuses when
It cannot fetch live sources. Never answers from training data or memory of past updates; a stale or hallucinated update name is worse than saying nothing.
Procedure, three stages, all mandatory
- What changed. Name, dates, confirmed or observed, what it targets. Sources are tiered: confirmed means the search engine's own status dashboard, official blog, or changelog; observed means volatility trackers or industry reporting. Every claim carries its tier.
- What it means here. Cross-reference
crawl.jsonfor which of this site's pages match the pattern the update targets,state.jsonfor their tiers,gsc.jsonfor position and impression deltas across the update window, andledger.jsonlfor what was shipped in that same window. Concrete counts, never general advice: "14 of your tier 1 pages match this pattern, average position moved from 8.2 to 11.4 across the window" beats "you may be affected." - What to do. Ranked by
revenue_score. Low-tier items are explicitly deprioritised with the reason shown, not silently dropped from the list.
Control check, mandatory
Report whether non-matching pages also moved in the same window. If they did, state plainly that the update is unlikely to be the cause. This is the same correlation-labelling rule as everywhere else in the system: a timing coincidence is not a demonstrated cause.
Decision rules
- Always fetch. Always stamp
fetched_aton the finding. - Never assert that the update caused a metric change without the control check alongside it.
- Write the
algo_updateevent regardless of whether the site was affected; a null result ("checked, not applicable here") is itself useful for later retrospectives.
Output
## What changed
<name>, <date range>, <confirmed|observed>, targets: <pattern>
## What it means here
<n> of our tier <n> pages match this pattern.
Average position: <before> to <after> across the window (gsc.json, <dates>).
Shipped in this window: <cause>, <n>h.
## What to do
<url> tier <n> score <x>: <action>
(deprioritised: <url> tier 4, low revenue impact)
## Control
Non-matching pages moved <n>% over the same window.
<Update is unlikely to be the cause. / Consistent with the update, not proven by it.>Done when
Every claim carries its source tier, the control check is present and answered either way, and an algo_update event was appended to the ledger so later retrospectives can attribute correctly.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

