Yandex Direct data-driven hypothesis planning skill
ya-direct-hypothesis-planning
A Codex skill turns already-served Yandex Direct campaign data into a prioritized backlog of testable hypotheses
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Can propose, and after approval carry out, budget and campaign changes
- Works with an advertising account's spend and conversion reports
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/ya-direct-hypothesis-planningDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a claude-code -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .claude/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .claude/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a cursor -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .agents/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .agents/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a github-copilot -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .github/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .github/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a codex -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .agents/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .agents/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a gemini-cli -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .agents/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .agents/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .devin/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .devin/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Formerly Windsurf.
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a cline -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .cline/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .cline/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a roo -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .roo/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .roo/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
A fork of Roo Code, same .roo folders.
Run in a terminal in the project folder
npx skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning -a opencode -yThe skills tool installs the current version from the repository. Add the -g flag to use the skill in every project.
Without third-party tools, from commit 68b14ca
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .agents/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .agents/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .agents/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .agents/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Run in a terminal in the project folder
tmp=$(mktemp -d)
git clone --filter=blob:none --no-checkout https://github.com/rdsolod-ui/ya-direct-hypothesis-planning.git "$tmp"
git -C "$tmp" sparse-checkout set --no-cone /skills/ya-direct-hypothesis-planning/
git -C "$tmp" checkout 68b14ca15b214b93bd80f13ed9f83d34794a74c2
mkdir -p .agents/skills
cp -R "$tmp/skills/ya-direct-hypothesis-planning" .agents/skills/ya-direct-hypothesis-planningCommands for macOS and Linux, on Windows run them in Git Bash.
Install with npx --yes skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning --agent codex --global --copy --yes, then supply a local delivery report and an explicit test budget envelope.
Other ways from the author
npx --yes skills add rdsolod-ui/ya-direct-hypothesis-planning --skill ya-direct-hypothesis-planning --agent codex --global --copy --yesInstall via the skills.sh registry.
This is third-party code. Review the repository files before installing.
What it does
The skill takes locally supplied reports and turns them into a hypothesis backlog: it first fixes the report contract together with the ya-direct-statistics skill, then normalizes rows, separating observed columns from calculated metrics, computes a weighted baseline and Wilson confidence intervals for CTR and conversion rate instead of averaging row-level ratios. It checks data sufficiency against user-set thresholds and returns INSUFFICIENT rather than a guessed conclusion when data is thin. It diagnoses where in the funnel a change occurs, builds one hypothesis per variable with an exact decision rule, and sizes a test budget only within an explicitly provided envelope. Exports, budget edits, campaign changes and launches stay behind explicit approval.
Who it is for. For people who already have trustworthy Yandex Direct delivery statistics and need it turned into a concrete test list, not vague conclusions.
Good fit when
- Campaigns already have impressions, clicks, spend and conversions, and you need to decide what to test next
- You need to tell a statistically real CTR difference from noise
- You need to size a test's budget and duration against an explicitly given amount
Not a fit when
- There is no delivery history yet: use the companion ya-direct-create-new-hypothesis skill
- Data does not meet your thresholds: the skill returns INSUFFICIENT, not a hypothesis
- You want the skill to launch the test or change budget itself: that stays behind separate approval
Example request
Here is a 60-day campaign report, build a hypothesis backlog with confidence intervals and propose a test budget within 20,000 rublesLimitations
Works only with locally supplied data; downloading or exporting advertiser data needs explicit approval at each step. Does not guess with an incomplete report contract, mixed attribution or unknown denominators, and honestly returns INSUFFICIENT instead. Install targets Codex via the skills.sh registry.
How to disable. Remove the skills/ya-direct-hypothesis-planning directory from your agent's installed skills.
Security check
- Can propose, and after approval carry out, budget and campaign changes
- Works with an advertising account's spend and conversion reports
README in short
The README describes the skill as a tool for hypotheses backed by already-served data and explicit test budgets, gives the Codex install command, and a ten-step workflow in SKILL.md from the report contract to preparing a handoff for the creatives skill. It separately lists local validation commands: validate_skill.py, unittest and script compilation.
SKILL.md
--- name: ya-direct-hypothesis-planning description: Analyze already-served Yandex Direct performance data and budgets to produce evidence-backed, prioritized, single-variable advertising hypotheses, test designs, budget proposals, and a creative handoff. Use when campaigns have historical impressions, clicks, spend, conversions, revenue, or element-level results. Fail closed on incomplete reporting contracts and keep exports, budget edits, campaign changes, and launches behind explicit approval. --- # Yandex Direct Hypothesis Planning Turn historical delivery data into a decision-ready test backlog. Do not treat correlation, platform recommendations, or small samples as causal proof.
FAQ
What does the skill do when data is thin?
It returns INSUFFICIENT instead of a hypothesis if the data misses the set sufficiency thresholds or attribution and denominators are undefined.
Can the skill launch the recommended test itself?
No, any export, budget edit, campaign change or launch needs separate explicit approval at that specific step.
Related
Official DataLens (Yandex) skills for Claude Code, Codex and OpenCode: SDK, HTML reports and RLS resolution
A Yandex Metrika MCP server generated from the API spec: 108 methods, 10 tools declared by default, transparent filters and response metadata
An open MCP server and agent skill set for SEO: keyword research, competitors, backlinks and site audits powered by DataForSEO
A Claude Code plugin for SEO audits: technical SEO, E-E-A-T, schema, local and AI search via parallel subagents