AgentSkillsCN

stash

将当前的上下文理解保存至 .claude/state。

SKILL.md
--- frontmatter
name: stash
description: Save current context understanding to .claude/state
disable-model-invocation: true
allowed-tools:
  - "Bash(~/.claude/skills/stash/scripts/prepare-stash.sh *)"

Stash Context Skill

Save your current understanding to .claude/state/context.md so you can resume later with a fresh context.

Workflow

  1. Analyze the current conversation to extract:

    • Problem summary: What we're working on (1-3 sentences)
    • Key decisions: Important choices made during the conversation
    • Relevant files: Files that matter and why
    • Current status: Where we are now
    • Next steps: What needs to happen next
  2. Base64 encode the content and pass as argument to the script:

bash
~/.claude/skills/stash/scripts/prepare-stash.sh 'BASE64_ENCODED_CONTENT'

The content to encode (before base64):

markdown
## Problem Summary
[1-3 sentences describing what we're working on]

## Key Decisions
- Decision 1
- Decision 2

## Relevant Files
- `path/to/file.ts` - why it matters

## Current Status
[Where we are now]

## Next Steps
1. Step 1
2. Step 2
  1. Tell the user: "Context stashed. Run /clear then /unstash to resume fresh."