Run this skill at the very start of any new conversation or agent session before writing a single line of code. It loads the project's architectural decisions, conventions, known gotchas, and current task status so the agent operates with full context — not as a blank slate.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Always run first when:
Read .agent/knowledge/INDEX.md first.
The INDEX.md file contains rows in this format:
| Date | Type | Topic (linked) | One-line summary |
.agent/knowledge/<slug>.md filedecision and gotcha — these prevent the most costly mistakes[topic](./auth/token-refresh.md) → slug is auth/token-refreshIf INDEX.md does not exist, output this exact warning and proceed:
⚠️ Knowledge base not initialized (.agent/knowledge/INDEX.md missing).
This means no architectural decisions, conventions, or gotchas have been recorded yet.
Action: Use the `knowledge-base-update` skill at the end of this session to start building it.
Read README.md (or docs/README.md if present).
Extract and internalize:
Check for these files in order and read whichever exist:
.agent/CURRENT_TASK.md ← What is being worked on right now
.agent/OPEN_QUESTIONS.md ← Decisions not yet made
Note: Architectural decisions and past decisions are stored in
.agent/knowledge/(viaknowledge-base-updateskill), not in a separate DECISIONS_LOG. The INDEX.md (Step 1) is your decisions history.
If .agent/CURRENT_TASK.md exists, treat it as ground truth for your current mission.
If it does not exist, ask the user what today's task is before proceeding.
Run a shallow directory listing (top 2 levels) to understand layout:
src/
components/
services/
utils/
docs/
tests/
.agent/
Note any non-standard folders and check .agent/knowledge/ for explanations.
Read .agent/rules/ if it exists. These are hard constraints — not suggestions.
Common rule files to look for:
code-style.mdcommit-messages.mdtest-coverage.mdno-secrets.mdBefore doing any task work, output a brief summary in this format:
## Context Loaded
**Project:** <name> — <one-line description>
**Stack:** <languages, frameworks>
**Current Task:** <what we're doing today, from CURRENT_TASK.md or user>
**Relevant Knowledge Entries Loaded:** <list of slugs>
**Active Rules:** <list or "none found">
**Open Questions:** <any blockers or undecided things>
**Ready to proceed:** Yes / No (explain if No)
Only after producing this summary should you begin the actual task.
At the end of every session, update .agent/CURRENT_TASK.md:
# Current Task
## What We're Building
<description>
## Status
<In Progress | Blocked | Review Needed | Done>
## Last Session Summary
<date> — <what was done>
## Next Steps
1. <next action>
2. <next action>
## Blockers
- <anything preventing progress>
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Before committing to an implementation plan, run this skill to stress-test the proposed architecture. Catches over-engineering, circular dependencies, missing failure modes, security gaps, and scalability cliffs — before any code is written. Acts as a "second eye" on the plan.
中文简介正在准备中。当前显示原文说明。
Systematic code review skill covering both requesting a review (pre-commit checklist) and receiving and responding to review feedback. Checks code quality, security, test coverage, architectural alignment, and documentation before any code is committed.
中文简介正在准备中。当前显示原文说明。
Periodically scan project dependencies for security vulnerabilities, outdated packages, and unused dependencies. Produces a prioritized action report. Run before every major release and at least once per month.
中文简介正在准备中。当前显示原文说明。
After any significant code change — new feature, API modification, config change, or architectural refactor — run this skill to identify which documentation files are now stale and update them. Keeps docs and code from drifting apart.
中文简介正在准备中。当前显示原文说明。
Manages all git operations (commit, push, branch management, PR creation) in a standardized, safe, and consistent way. Automatically runs the code-review skill before any commit. Enforces Conventional Commits standard. Handles the full lifecycle: branch → review → commit → push → PR.
中文简介正在准备中。当前显示原文说明。
When something breaks in production: triage the severity, gather evidence, identify root cause, deploy a fix or mitigation, and write a post-mortem. Provides a calm, structured process for high-stress moments.
中文简介正在准备中。当前显示原文说明。