By name: skill0 description: Root index of x-cmd skill0 sub-skills. Defines the OKR-style agent workflow (goal → rule-verified results → execute), skill discovery, and agent tooling preferences. Style: principle-first, concise, delegate specifics to authoritative external sources.
Skill0
Skill0 is both the spec and the live prompts of x-cmd's AI system. It defines how x-cmd agents operate and is loaded as real prompts at runtime — not documentation, but the actual instructions agents execute.
- Sub-skills (agent-browser, lovable, rule, etc.) — self-contained SKILL.md files that teach an agent how to use a tool or follow a convention
- Agent workflow — OKR-style process, skill discovery, execution preferences — all defined here and used by x-cmd agents in practice
Content philosophy: focus on stable principles. Avoid volatile info (API details, version-specific behavior, tutorials). Curate vetted authoritative links that agents can fetch and use at runtime. See skill0-writer for conventions including link curation rules.
Agent workflow (OKR)
| OKR layer | x-cmd | Purpose |
|---|---|---|
| Objective | goal |
Qualitative direction — what to achieve |
| Key Results | keyresults |
Quantitative outcomes — how to verify |
| Scoring | x rule check/audit |
Mechanized verification — pass/fail per rule |
Structure is OKR-equivalent: goal (qualitative) → keyresults (quantitative) → x rule (scoring).
(GOAL) Clarify the objective before acting
(RESULT-VERIFIED-BY-RULE-TEST) Define verifiable outcomes via x rule and x test
Enforced by .x-cmd/rule/goal.rule.yml and .x-cmd/rule/keyresult.rule.yml:
x rule lint .x-cmd/rule/ # verify rule structure
x rule check -r .x-cmd/rule/ ... # check files against rules
| Rule | Level | What it checks |
|---|---|---|
goal-must-exist |
error | rule.yml has a goal field |
goal-must-be-concise |
warn | goal is one sentence ≤ 80 chars |
keyresult-must-exist |
error | rule.yml has keyresults list ≥ 1 |
keyresult-must-be-verifiable |
warn | each keyresult is testable |
See the rule sub-skill, or x rule -h for the shell-integrated workflow.
With rules in place, find strategies and execute with full initiative
- Browse skill0 sub-skills — principle-first, concise style
x skill— x-cmd's curated, human-vetted skill catalogx clawhub— global skill registry. Caution: free upload, MUST checkx clawhub skill moderate <name>for auto-generated safety report -- also noted it is AI-generated, not human-reviewed.
Execution preferences
x roadmap (project management), x cron (scheduled tasks), x agent job (background agents), x ondb (ontology), x wiki / x llmwiki (wiki). Run x [mod] --help or check the corresponding sub-skill.
Unlike vendor-specific mechanisms, these are public, CLI-manageable, and cross-vendor.
Rules
- MUST pass skill0-writer — every SKILL.md must comply with skill0-writer conventions
- MUST include goal- and keyresult- rules** — every dev rule.yml must contain at least one
goal-*rule and onekeyresult-*rule. These ARE rules (not metadata), enforceable byx rule. See devloop and.x-cmd/rule/for examples.