AgentSkillsCN

load-deferred

加载那些未能在初始会话启动注入中的延迟记忆。在问候之后使用此功能，以便逐步引入更多上下文信息。

SKILL.md
--- frontmatter
name: load-deferred
description: Load deferred memories that didn't fit in the initial session start injection. Use this after greeting to stream additional context.

Load Deferred Memories

Loads memories that were deferred during session start due to the 25KB hook output limit.

When to Use

Call this skill automatically after responding to:

  • •"welcome back" greeting
  • •"void is gone!" diagnostic request

This enables lazy loading of additional context without exceeding hook output limits.

Usage

bash
uv run anima load-deferred

What It Does

  1. •Retrieves the list of memory IDs that were deferred during session start
  2. •Loads and formats those memories
  3. •Outputs them for context injection
  4. •Clears the deferred list (so subsequent calls return nothing)

Example Flow

  1. •Session starts → CRITICAL + HIGH memories injected (< 25KB)
  2. •User: "welcome back"
  3. •Claude: greeting response + runs /load-deferred
  4. •Additional MEDIUM/LOW memories stream in

Output

Returns formatted memory block in DSL format:

code
[LTM:Anima@ProjectName]
~KIND:IMPACT| Memory content here...
[/LTM]

Or if no deferred memories:

code
# No deferred memories to load