| name | context |
| description | Load relevant ADRs, patterns, and memory before implementation work. |
Context Skill
Use this skill before implementation to reduce rework and align with existing decisions.
Team-awareness
Read active team config from .codex-workflow/config/active-team.json when present.
Use its architecture_root and memory_root to locate context sources.
Modes
Quick mode
Use when the user asks for a fast orientation.
- Read
<memory_root>/memory-log.md.
- Read project overview context file(s).
- If a topic is provided, search memory for topic keywords.
- Return a short "recent activity + relevant references" summary.
Full mode (default)
- Infer work type from request:
new-endpoint, new-model, bug-fix, auth-change, testing
- Load relevant ADR summaries (limit to most relevant).
- Load relevant patterns for the inferred work type.
- Load related entity/context notes from memory.
- Return:
- work-type summary
- relevant ADRs
- relevant patterns
- checklist for implementation
Token discipline
- Prioritize summaries over long excerpts.
- Limit loaded references to what is necessary for the task at hand.