This document records the connections that were preserved when code-review was extracted into this repository.
The main skill depends directly on these in-repo resources:
skills/code-review/references/review-lanes.mdskills/code-review/references/review-procedures.mdskills/code-review/references/request-review-template.mdskills/code-review/references/security-best-practices/skills/code-review/references/rust-blockchain-critic/skills/code-review/agents/openai.yaml
These are separate skills, but they are part of the working graph for code-review:
skills/instruction-surface-audit/SKILL.md- used by
instruction-reviewworkflows and policy/glob/artifact coverage checks
- used by
skills/runtime-rollout-matrix/SKILL.md- used when review work also has rollout, lifecycle, replay, metrics, or multi-layer runtime risk
code-review-graph- not vendored here, but installed by
install.sh - expected by the
graphlane as an external CLI/tooling dependency - useful for blast-radius analysis and minimal review-set selection
- not vendored here, but installed by
These were intentionally documented rather than copied because they belong to the host workspace, not to the skill itself:
- host-level keyword routing that maps phrases like
code review,security review,cleanup, ordesloptocode-review - host
AGENTS.mdguidance that declarescode-reviewas the unified review surface - review-handoff or subagent prompts that reuse
request-review-template.md
The goal of this repository is to keep the code-review skill portable without pretending to be the whole original workspace. Everything above is either bundled here or explicitly called out so the host can reattach the missing integration points on purpose.