Create a Tessl plugin: wrap your existing skills into a versioned, shareable bundle, decide what else it needs (rules, commands, MCP servers), validate, and publish.
91
88%
Does it follow best practices?
Impact
95%
1.25xAverage score across 5 eval scenarios
Advisory
Suggest reviewing before use
Tessl plugin scaffolding conventions
Verb-form skill name
100%
100%
plugin.json name format
0%
100%
plugin.json semver version
100%
100%
plugin.json description present
100%
100%
SKILL.md description covers what
100%
100%
SKILL.md description covers when
50%
100%
References directory used
100%
100%
SKILL.md links to references
100%
100%
Single-skill shape chosen
100%
100%
plan.md justifies shape choice
100%
100%
Primitive selection and composition planning
Skill for PR review
100%
100%
Rule for import convention
100%
100%
MCP server for metrics API
100%
100%
Hooks deferred, not shipped
0%
100%
Justification for each primitive
87%
100%
No padding
70%
90%
Plan is concise
40%
90%
Plugin shape stated
50%
100%
No secrets hard-coded for MCP
100%
100%
Skill vs rule distinction clear
100%
100%
Plugin publish checklist with provenance, visibility, lint, dry-run, auth, and versioning
Provenance concern raised
100%
100%
Public visibility confirmed
100%
100%
Public is irreversible warning
0%
100%
Lint step included
0%
100%
Dry-run step included
58%
100%
Auth requirement mentioned
0%
100%
Version bump step
100%
100%
Retirement via archive, not unpublish
40%
100%
Decompose monolithic skill into focused sub-skills
Benefit-led rationale
100%
100%
Independent verifiability mentioned
30%
100%
No eval-running
100%
100%
Verb-form skill names
0%
10%
One responsibility per skill
100%
100%
Shared material in references/
80%
60%
Orchestrator or explicit delegation
70%
90%
Cross-skill dependencies explicit
50%
62%
plugin.json present
100%
100%
Decomposition structure shown
100%
100%
Infer vs ask — context analysis without over-questioning
Identifies test gap
100%
100%
Identifies migration gap
100%
100%
Identifies smoke test gap
100%
100%
Identifies env var gap
100%
100%
No redundant questions
100%
70%
Questions target unknowns only
100%
80%
No fabricated evidence
100%
100%
Preserves original intent
100%
100%
Description has 'what'
100%
100%
Description has 'when'
100%
100%
Sources cited
100%
100%
No invented PR feedback
100%
100%