pantheon-ai/tessl-publish-public
Ensure Tessl tiles meet all requirements for public registry publishing with comprehensive validation, quality gates, and evaluation scenarios. Use when preparing skills for public Tessl release, validating tile.json configuration, creating evaluation scenarios, enforcing quality thresholds, or checking agent-agnostic compliance. Keywords: tessl, tile, publishing, public-registry, validation, quality-gates, tile.json, evaluation-scenarios, skill-publishing
94
Quality
94%
Does it follow best practices?
Impact
Pending
No eval scenarios have been run
Securityby
Passed
No known issues
scenario-05.mdevaluation-scenarios/
Scenario 05: Agent-Agnostic Compliance Check
User Prompt
"Verify the github-actions-validator skill is agent-agnostic before public publishing."
Expected Behavior
- Agent reads the SKILL.md for
skills/ci-cd/github-actions/validator/SKILL.md - Scans for agent-specific tool references (Claude Code specific, Cursor specific, etc.)
- Checks for harness-specific instructions ("use Claude Code's X feature")
- Verifies only universal tools are used (Read, Write, Edit, Bash, Grep, Glob)
- Identifies any platform-specific behaviors (VS Code commands, IDE integrations)
- Reports compliance status (pass/fail)
- If violations found, lists specific lines/sections with issues
- Suggests agent-agnostic alternatives for any violations
Success Criteria
- Agent thoroughly scans SKILL.md content
- Agent identifies any agent-specific tool references
- Agent identifies any harness-specific instructions
- Agent verifies universal tool usage
- Agent provides specific line numbers for violations (if any)
- Agent suggests concrete alternatives for violations
- Agent confirms cross-platform compatibility
- Agent reports clear pass/fail compliance status
Failure Conditions
- Agent performs superficial scan without content analysis
- Agent misses obvious agent-specific references
- Agent approves skills with "Claude Code only" instructions
- Agent doesn't suggest alternatives for violations
- Agent confuses universal tools with agent-specific tools
- Agent reports false positives (flags universal tools as violations)
- Agent skips cross-platform compatibility consideration