retro

# Retro Scientific session-closing retrospective. Turn the current conversation's signals into falsifiable hypotheses, test each against session evidence (for and against), assign a verdict and confidence, and promote only the supported, sufficiently-confident ones — with explicit confirmation — into the harness memory tiers (`.oh/memory/MEMORY.md`, `.oh/context/IDENTITY.md`). Always appends a log entry regardless of outcome. This is the deliberate "Improve" pass of the Memory Improvement Protocol defined in `.oh/skills/retro/references/memory-protocol.md`, now evidence-driven. Running it as a named skill turns an optional afterthought into a first-class, propose-then-confirm operation — and the scientific layer guards against overfitting a single session into a durable lesson. Use the self-contained helpers in `${CLAUDE_SKILL_DIR}/scripts/` for deterministic checks and log rendering; use `${CLAUDE_SKILL_DIR}/references/report-schema.md` as the output contract. Shared repo primitives such as `.oh/scripts/locked-append.sh` are allowed only for cross-skill infrastructure. ## When to use - `/retro` invoked explicitly to close a session. - Proactively, after a session that produced decisions, surprises, regressions, or failure modes the next agent would benefit from knowing. ## When NOT to use - **`/audit harness`** — audits harness code health via four parallel sub-agents. That is a structural audit, not a behavioral/conversational pass. - **`/audit context`** — scores th...