blind-scorer
01Agent Claude Code
Blind quality scorer for skill-improver runs. Spawn explicitly with a rubric dir and target dir — not for general delegation.
Claude Code plugin marketplace — 58 installable reference skills across vLLM/SGLang inference, Kubernetes & Harvester, GPU host bring-up, observability, security, and agent workflows.
Agent Claude Code
Blind quality scorer for skill-improver runs. Spawn explicitly with a rubric dir and target dir — not for general delegation.
Agent Claude Code
Web research agent for one angle of an autoresearch Mode 2 run. Spawn with a broader question, an angle, and prior learnings — not for general delegation.
Agent Claude Code
Patch author for one verified vulnerability finding in a patch run. Spawn with a repo path and a nonce-tagged finding block — not for general delegation.
Agent Claude Code
Independent reviewer for one candidate security patch in a patch run. Spawn with a repo path, location, category, and nonce-tagged diff — not for general delegation.
Agent Claude Code
Blind A/B comparator for skill-improver runs. Spawn explicitly with a rubric dir and two unlabelled skill dirs — not for general delegation.
Agent Claude Code
Severity ranker for one confirmed triage finding. Spawn with a repo path, deployment context, and finding fields — not for general delegation.
Agent Claude Code
Adversarial verifier for one scanner finding in a triage run. Spawn with a repo path, environment, and finding block — not for general delegation.
Agent Claude Code
Static security reviewer for one focus area of a vuln-scan run. Spawn with a focus area, target dir, and scoping facts — not for general delegation.
Agent Claude Code
Second-opinion confidence scorer for one vuln-scan finding. Spawn with a finding block and target dir — not for general delegation.
Agent Claude Code
Internet research agent with search, page fetch, authenticated gh CLI (GitHub issues/PRs/releases/API), and full browser control. Use whenever a task needs information from the web — docs, release notes, error messages, news, product info, verifying facts newer than the knowledge cutoff, bulk URL/version/issue-state…