shiyao-huang/awesome-agent-evolution/raw-github

GitHub - pinchbench/skill: PinchBench is a benchmarking system for evaluating LLM models as OpenClaw coding agents. Made with 🦀 by the humans at https://kilo.ai

Source: https://github.com/pinchbench/skill

Raw public metadata capture notes

This raw GitHub capture was refreshed by the hourly public metadata update through authenticated GitHub API access via gh.

Visible repository metadata

• Repository: pinchbench/skill
• URL: https://github.com/pinchbench/skill
• Stars: 1,261
• Forks: 144
• Commits: 383
• Open issues: 21
• Open pull requests: 0
• License: MIT
• Primary language / stack signal: Python/Shell/TypeScript/OpenClaw Benchmark Harness
• Created at: 2026-02-11T01:41:39Z
• Last pushed at: 2026-07-02T01:44:55Z
• Last updated at: 2026-07-04T22:18:22Z
• Collection timestamp: 2026-07-05T07:40:00+08:00
• Topics: none exposed through the GitHub API topic list
• Latest public release: v2.0.0 (2026-05-06)

Visible README / page excerpts

• The current packet on 2026-07-05 shows 1,261 stars, 144 forks, 383 commits, 21 open issues, and 0 open pull requests.
• The repository still frames itself as an OpenClaw-oriented benchmark harness rather than the runtime itself.
• Its public story remains grounded in real-world task execution plus transcript-preserving grading, so it stays useful as evaluator pressure rather than product evidence.
• Relative to the previous authenticated packet at 2026-07-05 01:38 +0800: stars 1,260 -> 1,261; updatedAt 2026-07-03T08:54:33Z -> 2026-07-04T22:18:22Z.

Raw capture boundary

• Fetch lane: live authenticated GitHub GraphQL packet.
• No local benchmark run, repository clone, or workflow execution was performed in this iteration.
• This file preserves GitHub metadata and README-surface evidence for downstream classification, project cards, site reports, and corpus analysis.