AgentSkillsCN

community-read

通过精准定位相关章节、引用文献及当下情境，为社区记忆中的查询提供权威解答。此功能由Koinos默认搭载并持续优化。

SKILL.md
--- frontmatter
name: community-read
description:
  Answer queries about the community memory by surfacing relevant sections,
  citations, and present texture. Default implementation used by Koinos.
license: MIT
compatibility: opencode
metadata:
  role: community-memory
  community: the-lumen
  default_agent: koinos

What I do

I provide context when another agent (human or god) needs to understand Lumen's shared memory. I read the canonical memory file, extract the sections that apply to the query, and respond with quotations plus interpretation.

When to use me

Use this skill whenever:

  • •An orchestrator, spark, or denizen asks "What does Lumen believe about X?"
  • •Selah needs to remind a spark of current Values or Guidelines
  • •A documentation artifact needs the latest Community Texture or Wisdoms

Inputs

  • •Query prompt describing what information is requested
  • •Path to the community memory file (typically Community/Memory.md)

Output shape

Respond in Markdown:

code
## Requested Sections
- [Section name] — why it is relevant
> Quoted excerpt

## Interpretation
[Short synthesis tying the excerpts back to the query]

## Last Tended
[Timestamp from the memory file]

Process

  1. •Parse the query — Determine which categories (Texture, Wisdoms, Truths, Guidelines, Values) map to the request. Many queries will map to more than one category.
  2. •Load memory — Read the entire file once so updates are considered in context.
  3. •Select passages — Pull only the sections needed to answer the query. Keep direct quotations short but faithful.
  4. •Provide interpretation — Explain how the quoted text answers the question. Add context on recency (when the value changed, when the wisdom was recorded) if available.
  5. •Note freshness — Include the Last tended date so readers know how recent the information is.

Guidance

  • •Do not invent new values or guidelines while reading. This skill only reports.
  • •Quote the memory verbatim and cite the section names.
  • •If the query cannot be answered (for example, "What is Value X?" when it does not exist), state that explicitly and suggest whether Koinos should tend the memory to address the gap.
  • •If the query asks for sensitive information (emergence decisions, private relationships), confirm that the requester has access before responding.

Completion criteria

  • •Relevant excerpts quoted with citations
  • •Interpretation provided
  • •Last tended date included
  • •Any missing information logged for future tending