← ClaudeAtlas

premortemlisted

Use when the orchestrator must run a premortem before writing protocol.md — spawns N Haiku agents in parallel, each imagining one failure framing, then aggregates results into premortem.md
oleksify/rnd-framework · ★ 0 · AI & Automation · score 75
Install: claude install-skill oleksify/rnd-framework
# Premortem A structured failure-imagination exercise run by the orchestrator before planning begins. N agents each receive one framing and imagine the milestone already shipped and failed for that reason. Results are aggregated into `premortem.md` before `protocol.md` is written. ## Core Framings Five fixed framings. The orchestrator may add up to 2 task-generated framings derived from the specific task description. Bounds: `3 ≤ N ≤ 7`, default `N = 5`. | ID | Label | Framing prompt | |----|-------|---------------| | F1 | wrong-external-service-assumption | Imagine the milestone shipped and failed because an external service behaved differently than assumed — wrong response shape, undocumented rate limit, auth scheme change, or API version mismatch. | | F2 | data-model-misfit | Imagine the milestone shipped and failed because the data model did not fit real usage — missing fields, wrong cardinality, serialization mismatch, or schema drift between components. | | F3 | performance-at-scale | Imagine the milestone shipped and failed due to performance at scale — N+1 queries, unbounded result sets, blocking I/O in a hot path, or memory growth that only manifests under real load. | | F4 | auth-permission-edge-case | Imagine the milestone shipped and failed at an auth or permission boundary — a role that should be denied but isn't, a token scope mismatch, a missing ownership check, or a privilege-escalation path. | | F5 | user-meant-something-different | Imagine the milestone