What problem does it solve?
Teams using PostHog Review's automated PR reviews need custom review perspectives, blind-spot sweeps, validation bars, and resolution criteria tailored to their codebase, but authoring these LLMSkill rows requires knowing naming contracts, cardinality rules, and per-kind structure.
Core Features & Use Cases
- Four skill kinds covered: review perspectives (specialist lenses per PR chunk), blind-spot checks (final sweep conditioned on covered findings), validation criteria (publish/drop bar for findings), and resolution criteria (bar for implementing review threads).
- Naming and cardinality contracts: enforces prefixes like
review-hog-perspective-<slug> and rules such as exactly one active validation criteria per user.
- End-to-end authoring flow: ground in canonical skills via MCP skill tools, interview the user, draft the body, create the team LLMSkill with
posthog:skill-create, and explain activation in Inbox → Code review.
- Use Case: A team keeps getting burned by migration safety issues, so you author a custom
review-hog-perspective-migration-safety skill that hunts for unsafe schema changes in every PR chunk.
Quick Start
Ask the assistant to create a custom PostHog Review perspective focused on the concern your team cares about, and let it interview you, draft the body, and create the skill.
