# Benchmark Results Publisher > Publishes benchmark result JSONs into README/docs safely. Use when updating README tables, using MacBook RGB results, preserving Pillow results for the website, running tools/update_docs.sh, or preventing pyperf raw JSON from polluting documentation. Use when this capability is needed. - Skill: `tomevault-io/benchmark-results-publisher` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/benchmark-results-publisher` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/benchmark-results-publisher/raw - Safety review: pending (external: skill-scanner PASS, skillspector PASS) - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Docs & Writing - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/benchmark-results-publisher --- # Benchmark Results Publisher Use this when turning completed benchmark artifacts into documentation. ## Defaults While the full paper matrix is still running, `tools/update_docs.sh` defaults to MacBook M4 RGB micro results: ```bash ./tools/update_docs.sh ``` Equivalent explicit path: ```bash ./tools/update_docs.sh \ --image-results output/rgb_micro_macos_m4max/image-rgb/micro ``` Pass `--video-results` only when intentionally updating video README sections. ## Guardrails - Summary result files are `*_results.json`, including scenario names like `albumentationsx_micro_results.json`. - Raw pyperf files are `*.pyperf.json`; never load them as README columns. - RGB public docs should include Pillow when the result exists. - Do not commit `gcp_runs/`, `output/`, or `output_*/` artifacts. ## Validation Run after docs changes: ```bash python -m pytest tests/test_compare.py tests/test_update_readme.py pre-commit run --all-files ``` --- > Source: [albumentations-team/benchmark](https://github.com/albumentations-team/benchmark) — distributed by [TomeVault](https://tomevault.io).