Use to ingest a LongMemEval-style benchmark dataset into the isolated benchmark gramaton store via Claude Code sub-agents. Each unique haystack session becomes one Gramaton session via the production session_prepare → session_commit path. User-triggered by "run benchmark extraction", "ingest LongMemEval-S", "load benchmark data". Requires the gramaton-bench MCP server live on port 7338; see docs/benchmarks.md for setup.
$npx -y skills add gramaton-ai/gramaton --skill benchmark-extractInstalls into the current project.
Run `npx skills use "https://github.com/gramaton-ai/gramaton" --skill "gramaton-ai/gramaton/benchmark-extract"` and follow the generated skill instructions now. Read its complete output, redirecting it to a temporary file first if necessary. Resolve relative paths from the supporting-files directory it provides.
Use the skills in "https://github.com/gramaton-ai/gramaton" that are relevant to the current task. Run `npx skills add "https://github.com/gramaton-ai/gramaton"` and select the relevant skills, then follow their instructions.