token-budget plugins

by Melodic Software

Measures a Claude Code session's fixed startup context payload per item at a pinned binary version, including per-tool attribution of the built-in tool pools that /context reports only as totals, derived by A/B deny differencing under comparability rules. Falls back from an SDK meter to a headless /context parser, then to an error rather than a wrong number. Keeps a per-project before/after ledger per lever. Report-only by default; fix applies one approved project-scope trim.

by Wildcat Labs

Classify a repository's token sinks with evidence and emit the reading boundary agents respect.

by Matt Corbett

Prompt & context engineering team — agents (prompt-architect, prompt-implementation-engineer, prompt-reliability-engineer) that own the layer between the model and the app: prompt & context strategy (task decomposition, pattern selection across few-shot / CoT / decomposition / role, system-vs-user structure), the output-format contract (JSON mode, tool/function-calling schemas, few-shot curation, token budgeting), and prompt reliability (eval sets & regression, prompt versioning/CI, prompt-injection & jailbreak defense at the prompt layer). Provider-neutral. skills, a knowledge bank with Mermaid decision trees + a dated 2026 reference, best-practices, templates. Seams: which model → ai-coding-model-guidance / claude-api; retrieval → ai-rag-engineering; eval at scale → llm-evaluation-engineering; adversarial testing → ai-red-teaming; the app → claude-app-engineering / backend-engineering. Requires ravenclaude-core@>=0.7.0.

by trac3r00

66 agent skills across eight families: deterministic guards, interop, security, backend/deploy, domains, engineering process, power tools, and creative/frontend workflows, including audited Proxmox VE MCP administration.