stream-runtime-history

Implement or review bounded runtime-history storage and retrieval. Use for traces, sessions, artifacts, sensor summaries, replays, datasets, cursor pagination, JSONL streaming, pruning, retention, or any harness API whose cost can grow with historical records.

跳到安装

仓库: vinilana/dotcontext
最近来源活动: 2026年7月11日 15:53
检测到的 SKILL.md 语言: 英语
星标: 557
分支: 93

安装方式

默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

直接命令不会经过审查 Prompt；运行前请先检查来源。

npx skills add https://github.com/vinilana/dotcontext --skill stream-runtime-history

命令会保持在同一行。复制前请横向滚动并检查完整内容。

想先保存到本地？可下载 SkillsMP 当前能够提供的文件。

检查来源文件

决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。

文件资源管理器

2 个文件

SKILL.md

来源说明 · 只读预览

返回搜索返回安装

bound-mcp-payloads

Implement or review bounded MCP inputs and responses. Use for gateway response helpers, MCP Zod schemas, list actions, cursor pagination, maxEvents, JSON serialization, audit logging, response metadata, artifact exports, or payload-size and client-compatibility changes.

2026-07-11

bound-subprocess-output

Implement or review child-process execution with bounded stdout and stderr memory. Use for Node spawn wrappers, sensors, test runners, acceptance commands, timeout handling, output tails, truncation, or any change that captures subprocess output in the dotcontext harness.

2026-07-11

govern-cache-retention

Design, implement, or review bounded cache and persistent-index lifecycle. Use for ContextCache, SemanticContextBuilder, tree-sitter analysis cache, MCP session cache, hook host-session bindings, checkpoints, TTL, LRU, byte budgets, invalidation, cleanup timers, disposal, migration, or runtime retention configuration.

2026-07-11

查看这个仓库的全部 Skills

Stream Runtime History

Make memory and response size proportional to a requested page.

Workflow

Identify the cardinality and byte growth of every source collection.

Define default and maximum page sizes before changing an API.

Read JSONL incrementally and stop as soon as the page is complete.

Return an opaque, versioned cursor and explicit partial metadata.

Replace hot-path historical scans with bounded summary indexes.

Limit replay materialization at the source, transformation, response, and persistence layers.

Process multi-session work with bounded concurrency.

Add retention with dry-run, active-session protection, and crash-safe writes.

Required invariants

Never implement pagination by loading everything and slicing afterward.

Never return complete source arrays inside a partial replay.

Never use Promise.all over an unbounded set of sessions or files.

Keep point reads by ID available.

Read legacy runtime files without destructive migration.

Keep active workflow evidence immune from age-based pruning.

Treat malformed final JSONL lines as bounded diagnostics.

Cursor rules

Keep cursors opaque to callers, include a version, bind them to the query direction and filters, and return a typed error for invalid or stale values. Do not promise raw byte offsets as a stable public contract.

Dotcontext routing

Define pagination and retention in harness ports/application.

Keep MCP cursor adaptation thin.

Coordinate trace segments with the hook payload skill.

Follow F-03.

Review gate

Require load evidence for a large trace and many sessions. Verify page-size memory, maxEvents semantics, dataset concurrency, legacy reads, prune safety, and cursor boundary cases.

name	stream-runtime-history
description	Implement or review bounded runtime-history storage and retrieval. Use for traces, sessions, artifacts, sensor summaries, replays, datasets, cursor pagination, JSONL streaming, pruning, retention, or any harness API whose cost can grow with historical records.