Second Brain

Upgrade Blue CLI v0.6.6 → v0.7.0 in-session, refresh skill to v4.1, cascade pairing-contract + session-learnings + MCP backlog. CLI-first capability ownership for forms-domain authoring (copy, url, fi

2 min read

Upgrade Blue CLI v0.6.6 → v0.7.0 in-session, refresh skill to v4.1, cascade pairing-contract + session-learnings + MCP backlog. CLI-first capability ownership for forms-domain authoring (copy, url, fi

Decision

Upgrade Blue CLI v0.6.6 → v0.7.0 in-session, refresh skill to v4.1, cascade pairing-contract + session-learnings + MCP backlog. CLI-first capability ownership for forms-domain authoring (copy, url, field-level, chained-mutation create); MCP v4.0 retains LLM-driven CRUD; ~6 MCP tools added to v4.1 backlog (no immediate build).

Rationale

Five forces: (1) upstream release was a single feature commit (low blast-radius); (2) sha256-verified install + .bak preserved (Law-13 rollback); (3) cross-user verified pre-edit (openclaw + claude); (4) new CLI surface FILLED a real MCP gap (forms field-level ops), not a duplicated capability — Law-9 cohesion preserved; (5) skill is the SME — when CLI overtakes MCP in any domain, skill must route the matrix flip same operation, otherwise routing intelligence rots within hours. Bias audit (detect_biases) flagged anchoring on commit-count and framing (“only adds…“) — partially mitigated by the fact that the upgrade is reversible (binary swap) and that the cascade includes a documented MCP v4.1 backlog so the gap is tracked, not hidden. Honest gap: WRITE paths of new forms surface are untested against live API (only forms list was probed); confidence reflects that.

Alternatives Rejected

Outcome

Pending

Graph View

Backlinks