Skill Manifest Check
Paste a SKILL.md file -- the frontmatter-plus-Markdown format Claude Code and similar agent systems use -- and get every common authoring mistake flagged in seconds: missing or malformed frontmatter, a description too vague to trigger reliably, overlapping triggers between two or more skills, and a body-size caution against a context budget.
Structural pattern-matching, not a trigger test. This tool never runs your skill inside an agent and never calls an LLM to judge it -- it only looks at the text you paste and flags known failure shapes in it. A finding is something worth a look, not a verdict, and a clean result is not a guarantee the skill will trigger correctly.
1. Paste your SKILL.md
Paste one skill to check its frontmatter, description, and body size. Add a second (or more) to also check whether their descriptions overlap enough to compete for the same request. Nothing here is uploaded -- everything runs in this page.
Findings
Paste a SKILL.md above, or load the example, to see findings.
Overlapping triggers
Usage is optional, stays in this browser, and counts only that a lint ran or a summary was copied. Nothing you paste -- no SKILL.md contents, names, descriptions, or findings -- is ever recorded, uploaded, or sent to an analytics service. Uncheck to turn it off. Do Not Track and Global Privacy Control disable counting automatically.
Common questions
What does this tool check in a SKILL.md file?
Four things: the YAML frontmatter has a valid name and description, the description gives an agent enough to trigger reliably (length, a concrete trigger cue, and real distinctive vocabulary), whether two or more pasted skills have descriptions that could compete for the same request, and a rough token estimate for the instruction body against a documented caution line.
Does this tool contact a network, an AI model, or Anthropic's servers?
No. Every check is a deterministic pattern match against the text you paste, entirely in your browser. It never calls an LLM to judge the skill, never uploads anything, and never checks whether the skill actually triggers inside a real agent session.
Is a clean result a guarantee the skill will trigger correctly?
No. Every check here is a structural, pattern-matching signal against known failure shapes, not a semantic understanding of your description or a real trigger test. A skill with zero findings can still fail to trigger in an actual agent session, and a skill that trips a finding can still work fine -- read every finding as worth a look, not a verdict.
Why does pasting more than one SKILL.md matter?
An agent with many skills loaded has to pick one for each request. Two skills whose descriptions share too much distinctive vocabulary compete for the same request even if neither file is wrong on its own -- this tool only checks for that overlap when you paste two or more skills at once.