AgentSkillsCN

inject-context

在压缩之后，或当上下文丢失时，重新将冲刺上下文注入当前会话。此技能在每次波次之间，或当架构发生变动时，都显得尤为重要。

SKILL.md
--- frontmatter
name: inject-context
description: Re-inject sprint context into current session. Useful after compaction or when context is lost.
allowed-tools: Read, Glob
user-invocable: false

Context Injection

Re-injecting critical sprint context into your session.

Sprint Context

!cat .arch/sprints/sprint-*/CONTEXT.md 2>/dev/null | head -200 || cat .arch/sprints/current.md 2>/dev/null | head -100 || echo "No sprint context found"!

Active Wave Status

!cat .arch/sprints/sprint-*/WAVE-STATUS.md 2>/dev/null | head -100 || echo "No wave status found"!

Dependency Graph Summary

!cat .arch/sprints/sprint-*/DEPENDENCY-GRAPH.md 2>/dev/null | head -80 || echo "No dependency graph found"!

Your Agent Status

!cat .arch/agents/implementers/*.md 2>/dev/null | head -50 || echo "No agent files found"!

Current RFCs

In Progress

!ls .arch/rfcs/in-progress/ 2>/dev/null || echo "None"!

Approved (Available)

!ls .arch/rfcs/approved/ 2>/dev/null || echo "None"!

Context has been re-injected. Continue your work with the sprint context above.

If you are an implementer, remember:

  1. •Follow patterns from the CONTEXT.md above
  2. •Update WAVE-STATUS.md regularly
  3. •Use /claim-rfc and /complete-rfc commands

If you are an architect, remember:

  1. •Update context files when architecture changes
  2. •Document decisions as ADRs
  3. •Maintain comprehensive context for parallel agents