create-harness
Create or update an LLM harness that connects a language model to a kaggle-environments game (prompt generation, response parsing, and tests)
432 repo starsObserved in 1 repos
Agent Building
Agent Building skills
Browse reusable Agent Skills, each with a clear purpose and practical guidance.
run-ablation
Run a prompt-ablation study on a kaggle-environments game's LLM harness. Use when the user mentions "ablation", "prompt sensitivity", "test prompt variants", "compare prompts", "ablate the prompt", "prompt rewrite study", or asks whether their prompt wording is doing real work. Bootstraps a prompt_variants.py if missing, proposes new variants interactively, then runs a paired-seat tournament and reports the leaderboards.
432 repo starsObserved in 1 repos
Agent Building
team-interrupt
Send an interrupt signal to a working team agent. Use when an agent is actively running and you need to send it a message without having to wait until it has processed all other messages.
432 repo starsObserved in 1 repos
Agent Building
foundation-models-utilities
Use this skill when working with the `FoundationModelsUtilities` Swift package — a collection of utilities that extend Apple's Foundation Models framework with a chat completions client, on-demand "skills" that activate via tool calls, and history-management modifiers (drop completed tool calls, rolling window, summarization). Triggered when the user asks to "talk to a chat completions endpoint", "connect to OpenAI / a local LLM server", "add skills to a session", "manage transcript size", "summarize history", "drop tool calls", "rolling window", or works in a file that imports `FoundationModelsUtilities`.
431 repo starsObserved in 1 repos
Agent Building
agent-browser-issue
Open a concise GitHub follow-up for reusable browser-use limitations. Use when browser automation is blocked by a likely tool-side issue that is worth fixing separately, especially for clicks, dropdowns, file inputs, focus traps, or other repeatable agent/browser failures.
430 repo starsObserved in 1 repos
Agent Building
anchor-prior-test
Empirically test whether a candidate term qualifies as a semantic anchor by probing how densely it sits in LLM training data — across several model tiers in a clean room — before proposing it. Use when triaging an anchor proposal, deciding anchor vs contract, or vetting a rename. Produces a verdict against the four criteria (Precise, Rich, Consistent, Attributable), a tier rating, and a ready-to-paste proposal or rejection.
430 repo starsObserved in 1 repos
Agent Building
cc-usage-audit
한 프로젝트에서 사용자가 Claude Code에 입력한 프롬프트·작업 이력을 분석해 (1) 사용 패턴 정량화, (2) 사용자의 개발 철학 추출(근거 인용), (3) 그 철학을 렌즈로 한 메타 시스템(하니스·게이트·CI·프로세스) audit, (4) 우선순위가 매겨진 개선점 발굴을 수행한다. 트리거 — "내 프롬프트 분석해줘", "claude code 사용 패턴 분석", "내 개발 철학 추출", "메타 시스템 점검/audit", "하니스 개선점 발굴", "내가 입력한 프롬프트 기반으로 개선점 파악" 및 유사 의도.
430 repo starsObserved in 1 repos
Agent Building
harness-starter
A portable orchestrator that completes a high-level task through 5 stages — clarify → context-gather → plan → implement → verify. Each stage hands off via file artifacts, and a stage is skipped when its artifact already exists. The plan is decomposed into a task graph (DAG); independent nodes run in parallel, dependent nodes run in order.
430 repo starsObserved in 1 repos
Agent Building
memtrace-continuous-memory
Keep the Memtrace index fresh while editing by watching a repo for live, incremental re-indexing. Use when the user asks to keep Memtrace fresh while editing, watch a repo, enable live or incremental indexing, set up always-on memory (meaning Memtrace index watching, not generic agent memory), or make just-saved source code queryable immediately. Do not fall back to repeated Grep or manual rescans; configure Memtrace watching.
430 repo starsObserved in 1 repos
Agent Building
memtrace-fleet-first
Coordinate fleets of coding agents sharing one repo+branch: publish typed intents, classify edit episodes, and resolve conflicts before they collide. Use FIRST when multiple agents work the same repo+branch, before reading/planning/editing, when joining a fleet or handing work off, and when the user says two agents are changing the same thing, asks who should proceed, has a decision waiting, or asks you to mediate a Class C conflict. Covers branch-scoped publish-edit-record plus verdict, human-resolution, and directive polling. Do not grep for who else is touching a symbol or skip coordination because a change looks small. Skip only genuinely solo sessions or docs-only edits with no coordination value.
430 repo starsObserved in 1 repos
Agent Building