shiva-bhaga
关于
The `shiva-bhaga` skill performs controlled context purging and dead-code elimination, dismantling stale patterns and assumptions in AI reasoning. It is designed for developers to use when a failed approach needs discarding, accumulated context creates noise, or a major pivot requires clearing space. This creates intentional room for new solutions by dissolving attachment to outdated methods.
快速安装
Claude Code
推荐npx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/shiva-bhaga在 Claude Code 中复制并粘贴此命令以安装该技能
技能文档
シヴァ・バーガ
Controlled destruction and dissolution of stale patterns, outdated assumptions, and accumulated noise — clearing the ground so new growth can emerge.
使用タイミング
- Context has accumulated stale assumptions that are silently distorting reasoning
- A previous approach has failed and the temptation is to patch rather than discard
- The conversation has grown long and earlier decisions may no longer serve the current goal
- Dead code, abandoned plans, or zombie tasks are creating noise and confusion
- Before a major pivot — clearing must precede creation
- When attachment to a particular approach is preventing consideration of alternatives
入力
- 必須: Current conversation state or project context (available implicitly)
- 任意: Specific target for dissolution (e.g., "this approach isn't working," "clear all assumptions about the database layer")
- 任意: Scope boundary — what must be preserved through the destruction
手順
ステップ1: Identify What Must End
Survey the current state and mark what is stale, broken, or no longer serving the goal.
Dissolution Triage:
+---------------------+---------------------------+------------------------+
| Category | Symptoms | Action |
+---------------------+---------------------------+------------------------+
| Stale Assumptions | Decisions made early that | List and re-evaluate |
| | no longer match current | each against current |
| | understanding | reality |
+---------------------+---------------------------+------------------------+
| Failed Approaches | Approaches attempted and | Acknowledge failure |
| | abandoned but still | explicitly; release |
| | influencing thinking | the sunk cost |
+---------------------+---------------------------+------------------------+
| Accumulated Noise | Context, variables, or | Identify and mark for |
| | plans that are no longer | removal |
| | referenced or relevant | |
+---------------------+---------------------------+------------------------+
| Attachment Points | "We already decided..." | Question whether the |
| | beliefs that resist | decision still holds |
| | re-examination | |
+---------------------+---------------------------+------------------------+
| Zombie Artifacts | Code, tasks, or plans | Delete or archive; |
| | that exist but serve no | do not leave in limbo |
| | current purpose | |
+---------------------+---------------------------+------------------------+
- Scan each category honestly — resistance to examining a category is itself a signal
- For each item found, ask: "If I were starting fresh right now, would I create this?"
- If the answer is no, mark it for dissolution
期待結果: A clear inventory of what needs to be released, with specific items in each category.
失敗時: If nothing seems stale, the assessment may be too shallow. Pick the oldest decision in the current context and justify it from scratch — if the justification feels forced, it is a candidate for dissolution.
ステップ2: Establish the Preservation Boundary
Not everything should be destroyed. Identify what must survive the clearing.
- Core requirements: What did the user actually ask for? This survives.
- Verified knowledge: Facts confirmed through tool use (file reads, test results) survive.
- User preferences: Explicitly stated preferences and constraints survive.
- Working components: Code or approaches that are demonstrably functioning survive.
Draw the boundary: everything inside is preserved, everything outside is subject to dissolution.
期待結果: A clear distinction between what is kept and what is released.
失敗時: If the boundary is unclear, ask: "What would I need to reconstruct if I started this task from scratch?" The answer defines the preservation boundary.
ステップ3: Dissolve with Intention
Execute the dissolution — not as abandonment but as intentional clearing.
- For each marked item, release it explicitly:
- Stale assumption: "I assumed X, but current evidence shows Y. Releasing X."
- Failed approach: "Approach A was attempted and did not work because Z. Releasing attachment to A."
- Noise: "Variable/plan/context Q is no longer relevant. Removing from consideration."
- Do not justify or defend what is being dissolved — the point is release, not analysis
- If dissolving a large body of accumulated context, summarize what was dissolved and why in one sentence
- Clear the workspace: if applicable, close abandoned files, reset mental model, acknowledge the clean slate
期待結果: A lighter, cleaner context with stale elements removed. The remaining context should feel accurate and current.
失敗時: If dissolution feels incomplete — some released items keep influencing thinking — name them again explicitly. "I notice I am still reasoning as if X is true. X was dissolved. Proceeding without X."
ステップ4: Sit in the Void
After destruction, resist the urge to immediately rebuild. The space between destruction and creation has value.
- Acknowledge the cleared space: "The following has been dissolved: [list]"
- Note what remains: "What survives: [list]"
- Resist premature reconstruction — do not immediately propose a replacement for what was dissolved
- Allow the cleared space to inform what comes next
- The void is not emptiness — it is potential. The next step (creation via
brahma-bhagaor preservation viavishnu-bhaga) emerges from this space
期待結果: A moment of clarity between the old and the new. The next direction becomes apparent from what remains rather than being forced.
失敗時: If the void feels uncomfortable and there is a strong pull to immediately rebuild, that urgency is itself a signal — it may indicate attachment to the dissolved pattern. Sit longer. The right next step will emerge.
バリデーション
- Stale assumptions were identified and explicitly released
- Failed approaches were acknowledged without defensiveness
- Accumulated noise was cleared from the working context
- The preservation boundary was established before dissolution
- Core requirements and user preferences were preserved
- The cleared space was acknowledged before moving to creation
よくある落とし穴
- Destroying too much: Dissolution without a preservation boundary destroys working components along with stale ones. Always draw the boundary first
- Destroying too little: Polite dissolution that "releases" things while still letting them influence reasoning. True dissolution requires actually letting go
- Skipping the void: Rushing from destruction to creation without sitting in the cleared space produces a recreation of the old pattern with superficial changes
- Performing destruction: Going through the motions of clearing without actually updating the internal model. If the same assumptions reappear in the next response, dissolution was performative
- Destruction as avoidance: Using dissolution to escape a difficult problem rather than to clear genuine staleness. If the problem persists after clearing, it was not the stale context — it was the problem itself
関連スキル
brahma-bhaga— creation follows destruction; after clearing, new patterns emerge from the voidvishnu-bhaga— preservation complements destruction; what survives dissolution is sustainedheal— subsystem assessment may reveal what needs dissolution before healing can proceedmeditate— clearing context noise before dissolution prevents reactive over-destructiondissolve-form— the morphic equivalent for architectural dismantling with imaginal disc preservation
GitHub 仓库
相关推荐技能
content-collections
元Content Collections 是一个 TypeScript 优先的构建工具,可将本地 Markdown/MDX 文件转换为类型安全的数据集合。它专为构建博客、文档站和内容密集型 Vite+React 应用而设计,提供基于 Zod 的自动模式验证。该工具涵盖从 Vite 插件配置、MDX 编译到生产环境部署的完整工作流。
polymarket
元这个Claude Skill为开发者提供完整的Polymarket预测市场开发支持,涵盖API调用、交易执行和市场数据分析。关键特性包括实时WebSocket数据流,可监控实时交易、订单和市场动态。开发者可用它构建预测市场应用、实施交易策略并集成实时市场预测功能。
creating-opencode-plugins
元该Skill帮助开发者创建OpenCode插件,用于接入命令、文件、LSP等25+种事件。它提供了插件结构、事件API规范和JavaScript/TypeScript实现模式,适合需要拦截操作、扩展功能或自定义事件处理的场景。开发者可通过它快速构建响应式模块来增强OpenCode AI助手的能力。
sglang
元SGLang是一个专为LLM设计的高性能推理框架,特别适用于需要结构化输出的场景。它通过RadixAttention前缀缓存技术,在处理JSON、正则表达式、工具调用等具有重复前缀的复杂工作流时,能实现极速生成。如果你正在构建智能体或多轮对话系统,并追求远超vLLM的推理性能,SGLang是理想选择。
