lore-handoff

Lore Handoff

Use this skill when the next agent needs to pick up work safely without reading raw transcript soup.

Workflow

1. Use lore-recall to gather bounded evidence and status.
2. Reuse lore-brief/references/proposal-vocabulary.md for proposals, memory-card candidates, and contradictions.
3. Emit compact sections: verified, open, stale, risky, artifacts, proposals, memoryCardCandidates, contradictionCandidates, and nextActions.
4. Mark uncited/open claims explicitly; do not smooth uncertainty into certainty.
5. If evidence is stale, tell the next agent to inspect current files, tests, or runtime state before acting.

Rules

• Do not dump transcript blocks or page whole sessions.
• Every claim, artifact, and next action cites evidenceIds or is marked uncited:true.
• Preserve risky assumptions, unresolved questions, and contradictions with both sides cited.
• Use shared proposal, memory-card, and contradiction shapes; do not invent private todoList or mutation objects.
• Handoff may recommend next actions but does not perform them.

Validation

node skills/lore-handoff/scripts/validate-handoff.mjs skills/lore-handoff/examples/good-handoff.json
node skills/lore-dev-verification/scripts/validate-workflow-skill.mjs skills/lore-handoff