skill-creator-linter

name	skill-creator-linter
description	Internal tool (not user-invocable). Called by meta-skill-creator as a DAG step (kind: agent) to lint a candidate meta-skill SKILL.md against G1 (parse + reference check + xml_escape grep + structural lint) and G2 (scheduler dry-run with stub executors). Deterministic, sub-second, no LLM. Returns JSON diagnostics.
description_zh	内部工具（不可由用户直接调用）。由meta-skill-creator作为DAG步骤（kind: agent）调用，针对G1（解析+引用检查+xml_escape grep+结构lint）和G2（用桩执行器进行调度器空跑）对候选元技能SKILL.md进行lint。确定性、亚秒级、无LLM，返回JSON诊断。
user-invocable	false
disable-model-invocation	true
provenance	{"origin":"opensquilla-original","license":"Apache-2.0"}
metadata	{"requires":{"anyBins":["python","python3"]}}
entrypoint	{"command":"python {baseDir}/scripts/lint.py","args":["--skill-md-stdin","--gates","G1,G2"],"stdin":"{{ with.skill_md }}","parse":"json","timeout":30}

Skill Creator Linter

Validates a candidate meta-skill SKILL.md before it gets registered.

G1 rules

Rule	Check
G1.1	parse_meta_plan does not raise
G1.2	Every step.skill for kind=agent/skill_exec exists in the main catalog
G1.3	Template variables resolve at render time (covered by G1.1)
G1.4	on_failure parser rules (covered by G1.1)
G1.5	step kind: consistency (covered by G1.1)
G1.6	Grep: every `{{ inputs.user_message

G2 dry-run

Replace step executors with stubs yielding _StepDone(text="<stub:id>"). Run scheduler; pass if no exception and topology terminates.

Usage

uv run python {baseDir}/scripts/lint.py --skill-md path/to/SKILL.md --gates G1,G2
uv run python {baseDir}/scripts/lint.py --skill-md-stdin --gates G1,G2 < SKILL.md

Output

JSON to stdout: {"G1": {"passed": bool, "diagnostics": [...]}, "G2": {...}}.

Fallback

Manually inspect SKILL.md and run parse_meta_plan in a Python REPL.