<quick_start>
- •Check
./.{AGENT_NAME}/skills/memories/for memory skill folders.- •Each memory is a skill file at
./.{AGENT_NAME}/skills/memories/<name>/SKILL.md. - •Example baseline set:
- •
project-overview/SKILL.md - •
suggested-commands/SKILL.md - •
style-and-conventions/SKILL.md - •
task-completion-checklist/SKILL.md
- •
- •Each memory is a skill file at
- •If empty or missing, read
references/onboarding-guide.mdand execute onboarding. - •Otherwise, read memories and proceed.
- •If you learn new info, update the relevant
<name>/SKILL.mdvia$writing-memories. </quick_start>
<decision_points>
- •Memories missing/empty -> read
references/onboarding-guide.mdand run onboarding. </decision_points>
<quality_checklist>
- •Each memory should be dense with high-signal, repo-specific facts.
- •Prefer concrete evidence (paths, filenames, commands) over vague summaries.
- •If a memory is sparse, expand it by scanning the repo and updating it. </quality_checklist>
<failure_modes>
- •
AGENT_NAMEunclear: list directories under.and choose the active one (e.g..codex/or.claude/). </failure_modes>