Validate and atomically store immutable project activity events and receipts, then rebuild the derived activity index.
EnglishGenerate evidence-linked Guided Review artifacts and deterministic local Agent benchmark summaries. Use when reviewing Agent-produced diffs by intent, recording structured follow-ups, comparing multiple Agent/model/configuration runs against versioned repository assertions, or preparing reproducible review and benchmark evidence for /ship or Mission validation.
EnglishManages the visibility of the .agent directory in Git, resolving conflicts between plugin menu recognition and version control.
EnglishOptionally route branch-local plans, missions, handoffs, and incidents into collision-safe namespaces when clones or worktrees share one .agent tree.
EnglishClean temporary debug artifacts under .agent/debug (screenshots, logs, etc.) with age filters; optional .playwright-mcp cleanup.
EnglishA structured skill for evaluating the quality, reliability, and performance of a proposed code implementation. Invoked automatically when analyzing complex changes, refactoring, or introducing heavy dependencies.
EnglishManual and opt-in supervisor for the Cortex Agent Dashboard. Default disabled; auto enable makes the supervisor opt into lifecycle management.
EnglishTurn knowledge-health.json into low-risk documentation maintenance recommendations and emit doc-gardening-report.json.
EnglishOpenViking-style async self-evolution pipeline. After a session archive, worker classifies observations and writes them to .agent/memory/{feedback,project,experiences}/. v1 = deterministic, zero tokens.
EnglishResearch public GitHub repositories efficiently by using DeepWiki as a codebase map, then verify important claims against the repository's current primary sources. Use for architecture research, unfamiliar repository analysis, dependency evaluation, implementation comparisons, or when a user supplies a GitHub or DeepWiki repository URL.
EnglishRun deterministic knowledge checks for broken links, missing READMEs, plan lifecycle drift, and architecture doc mismatches, then emit knowledge-health.json.
EnglishLocal management query and controlled runtime-state writer for Cortex Agent dashboard, runs, queues, sessions, inbox, decisions, and waitpoints.
EnglishExternal resource ingestion (inspired by OpenViking's client.addresource). Pull URL / file / git repo into .agent/resources/external/, auto-generate L0/L1, register in context-index.json + uri-map.json + MANIFEST.json. Zero LLM dependency.
EnglishRecord and replay retrieval trajectories (inspired by OpenViking's observable retrieval). After every context-budget / knowledge-retrieval / skill-selector execution, writes .agent/runtime-evidence/trajectory/{task-id}{ts}.jsonl. Supports replay, URI resolution verification, and fixture export.
Englishsession-manager protocol plus Runtime Continuity v2 as a CLI: assess / log / checkpoint / archive / restore / status / warm / host-switch / resume-bundle / list-contexts. Writes transferable work state to .agent/runtime-continuity/ and readable archives to ~/.agent/contexts/ /.
EnglishAssess whether a feature is stateful or exempt, create and check its seven-part Runtime State Integration Contract, and summarize validation gaps. Use during architecture, planning, Mission, implementation review, or shipping of features with lifecycle, persistence, side effects, recovery, query surfaces, or durable evidence.
EnglishExpose one explicit Cortex Agent project through the standard, read-only Management API MCP stdio adapter.
EnglishDecide whether to open a Mission + Run based on session signals (turns, debug keywords, tool failures). Open, observe, record-tool. Keeps active-task persistence automatic per runtime-state-required.
EnglishReceiver for sub-agent fan-out lifecycle events. Hosts (Claude Code / Cursor / Codex) push spawned / progress / completed / cancelled via emit; this skill appends to runs/ .jsonsubagentfanout[] AND events[]. Use list / tree to reconstruct the fan-out tree.
EnglishResolve cortex:// URIs and .agent/ filesystem paths bidirectionally. Audits .agent/ markdown for relative references. Inspired by OpenViking's viking:// protocol. All new .agent/ documentation should prefer cortex:// form so cross-project copies don't require path rewrites.
EnglishCreate, check, and summarize executable validation contracts for Mission Lite milestones and high-risk tasks before implementation begins.
EnglishHarness 组件成熟度追踪。在 /ship CLEAN 阶段收集各组件表现指标,写入 metrics/component-health.json,供 /briefing…
ChineseVCS Pull Request 创建 / 状态 / 合并 — 通过 secret:// 间接 token,框架代为取 + 注入 Authorization header;agent 永不接触明文。可插拔 backend:gitea / github /…
Chinese
AGENT SKILLS ATLASIncludes Activity Recording, Agent Review Benchmark, Agent Visibility, and other workflows.
Claude Code
Codex
Cursor
GitHub Copilot
Gemini CLISupported by the Atlas install prompt. Skill-specific requirements vary.
Pack stats
What’s included
Activity Recording
Validate and atomically store immutable project activity events and receipts, then rebuild the derived activity index.
Agent Review Benchmark
Generate evidence-linked Guided Review artifacts and deterministic local Agent benchmark summaries. Use when reviewing Agent-produced diffs by intent, recording structured follow-ups, comparing multiple Agent/model/configuration runs against versioned repository assertions, or preparing reproducible review and benchmark evidence for /ship or Mission validation.
Agent Visibility
Manages the visibility of the .agent directory in Git, resolving conflicts between plugin menu recognition and version control.
Branch Namespace
Optionally route branch-local plans, missions, handoffs, and incidents into collision-safe namespaces when clones or worktrees share one .agent tree.
Cleanup Debug
Clean temporary debug artifacts under .agent/debug (screenshots, logs, etc.) with age filters; optional .playwright-mcp cleanup.
Code Evaluation
A structured skill for evaluating the quality, reliability, and performance of a proposed code implementation. Invoked automatically when analyzing complex changes, refactoring, or introducing heavy dependencies.
Dashboard Supervisor
Manual and opt-in supervisor for the Cortex Agent Dashboard. Default disabled; auto enable makes the supervisor opt into lifecycle management.
Doc Gardening
Turn knowledge-health.json into low-risk documentation maintenance recommendations and emit doc-gardening-report.json.
Evolution Pipeline
OpenViking-style async self-evolution pipeline. After a session archive, worker classifies observations and writes them to .agent/memory/{feedback,project,experiences}/. v1 = deterministic, zero tokens.
GitHub Repo Research
Research public GitHub repositories efficiently by using DeepWiki as a codebase map, then verify important claims against the repository's current primary sources. Use for architecture research, unfamiliar repository analysis, dependency evaluation, implementation comparisons, or when a user supplies a GitHub or DeepWiki repository URL.
Knowledge Lint
Run deterministic knowledge checks for broken links, missing READMEs, plan lifecycle drift, and architecture doc mismatches, then emit knowledge-health.json.
Management API
Local management query and controlled runtime-state writer for Cortex Agent dashboard, runs, queues, sessions, inbox, decisions, and waitpoints.
Resource Ingest
External resource ingestion (inspired by OpenViking's client.addresource). Pull URL / file / git repo into .agent/resources/external/, auto-generate L0/L1, register in context-index.json + uri-map.json + MANIFEST.json. Zero LLM dependency.
Retrieval Trajectory
Record and replay retrieval trajectories (inspired by OpenViking's observable retrieval). After every context-budget / knowledge-retrieval / skill-selector execution, writes .agent/runtime-evidence/trajectory/{task-id}{ts}.jsonl. Supports replay, URI resolution verification, and fixture export.
Runtime Continuity
session-manager protocol plus Runtime Continuity v2 as a CLI: assess / log / checkpoint / archive / restore / status / warm / host-switch / resume-bundle / list-contexts. Writes transferable work state to .agent/runtime-continuity/ and readable archives to ~/.agent/contexts/ /.
Runtime State Integration
Assess whether a feature is stateful or exempt, create and check its seven-part Runtime State Integration Contract, and summarize validation gaps. Use during architecture, planning, Mission, implementation review, or shipping of features with lifecycle, persistence, side effects, recovery, query surfaces, or durable evidence.
Runtime State MCP
Expose one explicit Cortex Agent project through the standard, read-only Management API MCP stdio adapter.
Session Triage
Decide whether to open a Mission + Run based on session signals (turns, debug keywords, tool failures). Open, observe, record-tool. Keeps active-task persistence automatic per runtime-state-required.
Subagent Trace
Receiver for sub-agent fan-out lifecycle events. Hosts (Claude Code / Cursor / Codex) push spawned / progress / completed / cancelled via emit; this skill appends to runs/ .jsonsubagentfanout[] AND events[]. Use list / tree to reconstruct the fan-out tree.
Uri Resolver
Resolve cortex:// URIs and .agent/ filesystem paths bidirectionally. Audits .agent/ markdown for relative references. Inspired by OpenViking's viking:// protocol. All new .agent/ documentation should prefer cortex:// form so cross-project copies don't require path rewrites.
Validation Contract
Create, check, and summarize executable validation contracts for Mission Lite milestones and high-risk tasks before implementation begins.
Maturity Tracker
Harness 组件成熟度追踪。在 /ship CLEAN 阶段收集各组件表现指标,写入 metrics/component-health.json,供 /briefing 展示成熟度看板和退化决策依据。
Vcs PR
VCS Pull Request 创建 / 状态 / 合并 — 通过 secret:// 间接 token,框架代为取 + 注入 Authorization header;agent 永不接触明文。可插拔 backend:gitea / github / gitlab。
Quick start
- Choose your skills. Select individual skills above, or choose the entire pack.
- Copy the install prompt. Use Copy install prompt above the list.
- Send it to your agent. Paste into your coding agent’s chat. Your agent reviews the source and guides the setup.
Try it with your agent
After installation, start with a request like:
Use the Activity Recording skill to help with my task.
Ask me for the context you need before getting started.Requirements & license
- A coding agent that can access your project files and install skills.
- Your agent checks each skill’s instructions, dependencies, and permissions during setup.
- Reported license for the skills shown: MIT. The author’s terms apply.