Skip to content

cli-reference.md

Latest commit

 

History

History
112 lines (79 loc) · 4.91 KB

cli-reference.md

File metadata and controls

112 lines (79 loc) · 4.91 KB
Raw

CLI Reference

skillranger is a single Node.js binary (npx -y skillranger@latest or a local install). Run skillranger --help for the full command list; skillranger <command> --help shows one command's options.

Requires Node.js 20 or newer.

Quick start command

# Interactive setup: scan, recommend, audit, plan, confirm, install
npx -y skillranger@latest setup

Inspecting projects and recommendations

# Detect repository context
npx -y skillranger@latest scan .

# Explain recommendations for a task
npx -y skillranger@latest recommend . \
  --target codex \
  --intent "Review this Next.js app before release" \
  --explain

# Audit one package
npx -y skillranger@latest audit frontend.next-app-router-review

Installing skills

# Preview installation (writes nothing)
npx -y skillranger@latest install frontend.next-app-router-review \
  --project . \
  --target codex \
  --scope repo \
  --dry-run

# Apply the reviewed plan
npx -y skillranger@latest install frontend.next-app-router-review \
  --project . \
  --target codex \
  --scope repo \
  --yes

# Inspect installed skills
npx -y skillranger@latest installed .

# Verify lockfile and file integrity
npx -y skillranger@latest verify .

# Remove a skill and update the lockfile
npx -y skillranger@latest uninstall frontend.next-app-router-review --project . --target codex

setup accepts the same flags as install (--target, --scope, --copy, --yes, --intent, --lane, --limit-per-lane) and additionally --no-agent-context to skip the AGENTS.md block for repo scope.

Core (universal) skills (core.proportional-engineering, core.universal-safety) are not project-specific: install them at user scope (--scope user) so one lockfile covers all projects. Repo scope remains an option for team pinning, and user-scope installation is sufficient for strict-mode lockfile matching.

Task routing

The task family prepares a natural-language task into a bounded, reviewable skill set. Two paths are supported: a deterministic fallback that matches the prompt and repository signals against a bounded bilingual vocabulary, and catalog-assisted routing where an MCP host model reads the audited catalog and proposes a prompt-grounded set — which SkillRanger validates before composing the final set. A model cannot add arbitrary skills, bypass audit or compatibility checks, or force an ineligible skill.

# Direct task routing (structured result, no interactive flow)
npx -y skillranger@latest task . \
  --intent "Review accessibility and fix critical focus traps" \
  --target codex \
  --json

# Read required instruction chunks from a routed task
npx -y skillranger@latest task:read . \
  --router-run <router-run-id> \
  --mandatory-next \
  --expected-read-revision 0 \
  --json

Router configuration

Routing limits live in the project skillranger.config.json (router object, schema schemas/router-config.schema.json): maxEnvironmentSkills, maxTaskCompanions, maxVerificationSkills, maxAgentContextSkills, maxCoreSkills (default 3, range 0..4 — the always-on core universal skills included in every prepared run), maxTotalSelectedSkills, maxInstructionBytes, maxAdditionalReadBytes, maxSingleFileBytes, maxIntentBytes, maxSelectedRisk, enabled, and strictByDefault. Core skills count toward the instruction-byte budget but never toward the total-skill cap or the agent-context slot.

Persisted runs

run:* commands start, step, evidence, verify, and finalize persisted runs. Runs track skill reads, execution, evidence, and verification; a run is only verified when every declared gate passes with attached evidence. See docs/workflow-runtime.md and docs/verification-engine.md for the full lifecycle.

Registry and release gates

npx -y skillranger@latest validate:registry   # manifest schema validation
npx -y skillranger@latest lint:skills         # skill content lint
npx -y skillranger@latest audit:registry      # static audit of every bundled skill
npx -y skillranger@latest publish:check       # validation + audit publication gates
npx -y skillranger@latest release:validate    # frontend release artifact contract

Design pipeline commands

design:* commands create briefs, recommend recipes, run browser observations, validate sources, verify artifacts, and produce bounded repair requests. See docs/design-rule-library.md, docs/browser-adapter.md, and docs/visual-benchmark.md.

Domain packs

domain:list and domain:inspect show the installed domain packs and their capabilities. See docs/domains/.

MCP server

npx -y skillranger@latest mcp

Starts the stdio MCP server; configure it in the host client per docs/mcp-host-config.md. skillranger doctor prints runtime and registry diagnostics.