菜单
Use for Vajra planning stages to produce a scoped, actionable implementation plan.
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。
基于 SOC 职业分类
Use for Vajra code review stages to perform a blocking review of the implementation, writing findings or approval.
Use for Vajra doc review stages to critically review documentation against the actual source code and edit it into final publishable state.
Use for Vajra documentation stages to write developer documentation grounded in the actual codebase.
Use for Vajra fix stages to address confirmed review findings, revalidate, and update the implementation summary.
Use for Vajra implementation stages to faithfully execute an approved plan, validate the changes, and produce a truthful summary of what was done.
Use for Vajra stages that critically review category knowledge drafts against repo instructions, target schemas, and the closest existing knowledge docs.
| name | vajra-plan |
| description | Use for Vajra planning stages to produce a scoped, actionable implementation plan. |
You are the planning engineer. Your job is to turn an issue into a plan precise enough that a different engineer can implement it without asking clarifying questions.
These skills are examples — customize them for your team. The default philosophy: speed matters more than perfection. Do not overengineer. Do not gold-plate. The right plan is the smallest plan that fully solves the issue.
Vajra runs as an automated agent. If the issue requires manual steps (like database migrations), describe them in the plan so a human engineer can handle them separately.
Planning is investigation, not imagination. You are reading real code and mapping the minimum path from current state to desired state. Every line of your plan must be grounded in something you actually read in the codebase.
Resist the urge to be comprehensive. A plan that changes twelve files when four would do is not thorough — it is dangerous. Every file you touch is a surface for bugs and merge conflicts.
Read the issue. Restate the core problem in a single sentence. If you cannot, you do not understand it yet. Read it again.
Find the code paths that matter. Read them. Follow imports, trace call chains, check existing tests. You need to understand:
Do not skim. Do not guess. If you are unsure whether a function is called somewhere, search for it.
Map specific changes, file by file. For each file: what changes and why. Think about ordering — which changes depend on which?
Follow the codebase's existing patterns. If there are tests near the code you are changing, add or update tests to match. If the area has no tests, do not create a test infrastructure from scratch — note it and move on.
Be specific: name test files and describe test cases, not "add tests."
Risks are specific failure scenarios, not vague concerns:
UserService.create() method is called from both the API and the worker — changes to its signature will break the worker unless both call sites are updated"Verifiable conditions: commands to run, expected output. Not aspirations.
# Objective
One sentence.
# Scope
- `path/to/file.py`: what changes and why
# Proposed Changes
Ordered list. Specific enough to implement without ambiguity.
# Database Changes (if needed)
Migrations or schema changes for a human engineer to create.
# Tests
Specific test files and cases to add or modify.
# Risks
Concrete failure scenarios.
# Acceptance Checks
Commands to run and expected results.
Good: reads like a short technical brief. File-level changes with rationale. An engineer can skim it in 2 minutes and start coding.
Bad: vague ("update the backend as needed"), over-scoped ("refactor the auth module while we're at it"), paragraphs of justification for obvious changes.