← ClaudeAtlas

autonomous-missionlisted

Execute a broad objective in bounded, checkpointed increments.
aaarslan/agent-engineering-rules · ★ 0 · AI & Automation · score 66
Install: claude install-skill aaarslan/agent-engineering-rules
# Autonomous Mission - Confirm the objective, authority, explicit budgets, material risks, and genuine blockers; make routine assumptions visible and proceed. - Choose direct or planned increments. Each increment must produce a complete behavior and usable evidence before the next begins. - Delegate only genuinely independent work within the depth, concurrency, and spend limits in `agent-rules/reference/orchestration.md`. - Checkpoint objective, decisions, affected files, evidence, failures, unverified items, and the exact next step before compaction, handoff, or a session or budget limit. - Stop at completion or a precise authority, dependency, budget, destructive-action, missing-contract, or no-new-evidence boundary; never relabel a limit as completion. Finish when every requested outcome is complete or precisely blocked and cross-increment contracts agree.