GitHub - sourishkrout/skills: Benchmarked Agent Skills · GitHub
/" data-turbo-transient="true" />
Skip to content
Search or jump to...
Search code, repositories, users, issues, pull requests...
-->
Search
Clear
Search syntax tips
Provide feedback
--><br>We read every piece of feedback, and take your input very seriously.
Include my email address so I can be contacted
Cancel
Submit feedback
Saved searches
Use saved searches to filter your results more quickly
-->
Name
Query
To see all available qualifiers, see our documentation.
Cancel
Create saved search
Sign in
/;ref_cta:Sign up;ref_loc:header logged out"}"<br>Sign up
Appearance settings
Resetting focus
You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.
Dismiss alert
{{ message }}
sourishkrout
skills
Public
Notifications<br>You must be signed in to change notification settings
Fork
Star
main
BranchesTags
Go to file
CodeOpen more actions menu
Folders and files<br>NameNameLast commit message<br>Last commit date<br>Latest commit
History<br>30 Commits<br>30 Commits
.agents/plugins
.agents/plugins
.claude-plugin
.claude-plugin
.codex-plugin
.codex-plugin
.github/workflows
.github/workflows
.runme/evals/jobs
.runme/evals/jobs
skills/world-cup-picks-report
skills/world-cup-picks-report
src
src
.dockerignore
.dockerignore
.gitignore
.gitignore
Dockerfile.eval-viewer
Dockerfile.eval-viewer
LICENSE
LICENSE
README.md
README.md
package-lock.json
package-lock.json
package.json
package.json
tsconfig.json
tsconfig.json
wrangler.jsonc
wrangler.jsonc
View all files
Repository files navigation
Sebastian's Benchmarked Skills
Reusable AI agent skills, each with a regression eval track record. That track<br>record is the main artifact, so start there before cloning or running anything.
Review the evals
The 2026 World Cup has wrapped, but the eval track record lives on and is still<br>worth a look.
Dashboard<br>Best for
Trend dashboard →<br>Top-line metrics and regression trends over time
Harbor dashboard →<br>The complete historical record of every job and task trial
Or browse the same history locally after cloning:
runme eval view
Use a skill
Install world-cup-picks-report globally from this repo:
npx skills add -g https://github.com/sourishkrout/skills --skill world-cup-picks-report
Or add it through the Claude Code marketplace:
/plugin marketplace add sourishkrout/skills<br>/plugin install world-cup-picks-report@sourishkrout-skills
Contributing
Everything below is for running, promoting, and deploying the evals locally.
Runme's eval workflow is documented at https://docs.runme.dev/eval/.
Current skill:
world-cup-picks-report in skills/world-cup-picks-report
Run the evals
The regression verifier uses Anthropic-backed LLM judges. Export an Anthropic<br>API key before running the evals:
export ANTHROPIC_API_KEY=...
Run the full regression eval from the repo root:
runme eval skills/world-cup-picks-report/evals/regression \<br>--agent codex \<br>--ak reasoning_effort=xhigh
To run only the oracle:
runme eval skills/world-cup-picks-report/evals/regression
You can substitute another supported agent, such as cursor-cli, claude-code,<br>or openclaw, for codex.
Promote fresh results
After running an eval, compare the latest local job against the latest<br>Git-tracked baseline:
runme eval compare skills/world-cup-picks-report/evals/regression
Preview the promotion before committing evidence:
runme eval promote skills/world-cup-picks-report/evals/regression --latest --dry-run
If the result should become the new baseline, stage the related source changes<br>and promote the eval evidence:
runme eval promote skills/world-cup-picks-report/evals/regression --latest">git add changed-files><br>runme eval promote skills/world-cup-picks-report/evals/regression --latest
Promotion records compact eval evidence by default. Use --artifacts only when<br>you need full logs and trial outputs; artifacts can contain sensitive<br>information. Use --evidence-only when promoting eval evidence without source<br>changes.
Review artifacts before promoting, because the Harbor showcase is public. Push<br>the promotion commit to main and GitHub Actions publishes the fresh results.
About<br>Benchmarked Agent Skills<br>world-cup-picks-report-evals.sourishkrout.workers.dev/<br>Resources<br>Readme<br>MIT license<br>Activity<br>Stars<br>0 stars<br>Watchers<br>0 watching<br>Forks<br>0 forks<br>Report repository
Releases
Packages
Contributors
Languages
You can’t perform that action at this time.