Commands, agents, and skills
This page lists what the plugin distributes. The distribution manifests are .claude-plugin/plugin.json (Claude Code) and .codex-plugin/plugin.json (Codex).
| Kind | Claude Code | Codex |
|---|---|---|
| Slash commands | Distributed | Not in the manifest |
| Sub-agent | Distributed | Not in the manifest |
| Skills | Distributed | Distributed |
.codex-plugin/plugin.json declares only skills and interface; it does not declare commands or agents.
Commands (Claude Code)
The seven entries registered under commands in .claude-plugin/plugin.json. Descriptions summarize the frontmatter description of each commands/*.md.
| Command | What it does |
|---|---|
/river-review:setup-team | Create .river/rules.md, confirm the plugin install, and check the integration mode |
/river-review:review-local | Review uncommitted changes, or the branch diff against the default branch when nothing is uncommitted |
/river-review:review-team | Parallel multi-role review merged with consensusLevel and a Tech Lead report |
/river-review:challenge | Pre-mortem, war-game, and logic-torturing analysis on current changes |
/river-review:skill | Load the best-matching skill doc from skills/ and apply it to the current task |
/river-review:check | Pre-review check: run lint/tests and summarize failures and risks |
/river-review:pr | Draft a review-ready PR description from a template |
Agent (Claude Code)
| Agent | Definition | What it does |
|---|---|---|
river-review | agents/river-review.md | Follows the river-review skill, runs perspective-based reviewer roles, and verifies findings against the diff |
Skills (Claude Code / Codex)
The 12 skills under skills/agent-skills/. Descriptions summarize each SKILL.md description.
| Skill | What it does |
|---|---|
river-review | Main entry point: intent classification, specialist skill selection, execution, and finding verification |
river-review-code | General code quality; the default fallback |
river-review-security | Regular security review (authn/authz, privacy design, and more) |
river-review-security-audit | Explicit security audit of a repository or subsystem |
river-review-performance | N+1 queries, memory efficiency, caching strategy, observability |
river-review-architecture | Dependencies, boundary design, data models, API design |
river-review-testing | Test coverage, naming, flaky tests, coverage gaps |
river-review-frontend | Accessibility, design-system conformance, UI state design, and more |
river-review-docs | Consistency between README / docs / AGENTS.md and the implementation |
adversarial-review | Adversarial review with Pre-mortem / War Game / Logic Torturing |
review-team | Runs six specialist reviewer roles in parallel and merges the results |
unknown-coverage-review | Synthesizes unverified items (Unknowns) left in a diff and its verification evidence |
In Claude Code, skills can also be invoked as /river-review:<skill-name>.