Skip to content

docs(readme): lead the pr-af card with its Code-Review-Bench result - #877

Open
oktaygoktas wants to merge 1 commit into
mainfrom
docs/readme-pr-af-hero
Open

docs(readme): lead the pr-af card with its Code-Review-Bench result#877
oktaygoktas wants to merge 1 commit into
mainfrom
docs/readme-pr-af-hero

Conversation

@oktaygoktas

Copy link
Copy Markdown
Collaborator

Summary

The Built With AgentField card for pr-af was showing the older "Open-Source Agentic PR Review" architecture diagram. This swaps it for the repo's current hero image — the Code-Review-Bench chart — and brings the card's copy and link in line with the rest of the grid.

  • New imageassets/examples/agentic-pr-reviewer-hero.png, taken from Agent-Field/pr-af's assets/hero.png. Same 16:9 aspect ratio as the image it replaces (1376×768 vs 2752×1536), so the two-column grid renders unchanged, and it's 1.0 MB instead of 4.6 MB.
  • Previous image keptassets/examples/agentic-pr-reviewer.png is untouched and stays in the repo.
  • Caption rewritten to lead with the result the image states, then keep the mechanism in one clause.
  • Link normalized from agentfield.ai/pr-af/ to agentfield.ai/github/pr-af/, the form every other card uses. utm_id is unchanged (github-readme-pr-af-repo), so analytics continuity is preserved.
  • assets/utm-links.csv — pr-af was the only "Built with AF" card missing from the manifest; added its row.

Caption

Before:

Builds a custom review strategy for every PR - spawns parallel reviewer agents with runtime-crafted prompts, adversarially challenges its own findings, and posts evidence-grounded inline comments.

After:

#1 open-source reviewer on Code-Review-Bench - 0.706 golden recall across 42 tools compared, at ~10x lower cost per review. Builds a custom review strategy for every PR, then adversarially challenges its own findings.

Both figures come from pr-af's own README ("PR-AF with GLM-5.2 is the #1 open-source reviewer in golden recall: 0.706 across 42 compared tools") and the image ("About 10x cheaper per review than closed-source tools"). The claim is scoped to open-source deliberately — cubic-dev scores 0.741 and is closed-source, so an unqualified "#1" would be inaccurate.

Verification

  • Both agentfield.ai/pr-af/ and agentfield.ai/github/pr-af/ return 200 and resolve to https://github.com/Agent-Field/pr-af, so the link change breaks nothing.
  • assets/examples/agentic-pr-reviewer.png had exactly one reference in the repo (the line changed here); no other file points at it.
  • Image stored under assets/examples/ as kebab-case PNG, matching the convention used by all six cards.

Note

assets/utm-links.csv remains incomplete relative to the README overall — only the pr-af row is added here, since that's what this change touches.

🤖 Generated with Claude Code

Swap the pr-af card image for the repo's hero chart (#1 open-source on
Code-Review-Bench) and rewrite the caption to match. Normalize the link
to the /github/<slug>/ form used by every other card, and add the
missing pr-af row to assets/utm-links.csv.

The previous image is kept at assets/examples/agentic-pr-reviewer.png.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant