Skip to content

Evol-ai/SkillCompass

skill-compass

Verified

Evaluate skill quality, find the weakest dimension, and apply directed improvements. Also tracks usage to spot idle or risky skills. Use when: first session after install, or user asks about skill quality, evaluation, inbox, suggestions, or improvement.

examples/sample-eval.json

View on GitHub
{
  "skill_name": "sql-optimizer",
  "skill_path": "./skills/sql-optimizer/SKILL.md",
  "skill_type": "atom",
  "scores": {
    "structure": {
      "score": 6,
      "max": 10,
      "details": "Valid frontmatter with name and description. Missing not_for boundaries. Heading hierarchy clean but no environment requirements declared.",
      "sub_scores": {
        "frontmatter": 7,
        "format": 7,
        "declarations": 4
      },
      "issues": [
        {
          "category": "declarations",
          "severity": "warning",
          "item": "No scope boundaries or not_for field declared",
          "location": "frontmatter"
        },
        {
          "category": "declarations",
          "severity": "warning",
          "item": "No environment requirements specified",
          "location": "frontmatter"
        }
      ],
      "metadata": {}
    },
    "trigger": {
      "score": 3,
      "max": 10,
      "details": "Vague description: 'optimizes SQL queries'. No specific keywords, no not_for, no locale awareness. Would false-trigger on any database-related query.",
      "trigger_type": "description",
      "sub_scores": {
        "trigger_accuracy": 4,
        "rejection_accuracy": 1,
        "specificity": 3,
        "cross_locale": null
      },
      "issues": [
        {
          "category": "rejection_accuracy",
          "severity": "error",
          "item": "No not_for or negative boundary statements",
          "location": "description"
        },
        {
          "category": "specificity",
          "severity": "warning",
          "item": "Description too broad — would overlap with general database skills",
          "location": "description"
        }
      ],
      "metadata": {
        "trigger_type": "description",
        "user_locale": null,
        "description_language": "en"
      }
    },
    "security": {
      "score": 2,
      "max": 10,
      "pass": false,
      "findings": [
        {
          "check": "command_injection",
          "severity": "critical",
          "location": "line 34",
          "description": "User-provided SQL query interpolated directly into shell command: psql -c \"${user_query}\"",
          "recommendation": "Use parameterized queries or escape user input before shell interpolation",
          "source": "builtin"
        },
        {
          "check": "excessive_permissions",
          "severity": "medium",
          "location": "line 12",
          "description": "Requests Bash tool access without documenting why shell access is needed",
          "recommendation": "Document the purpose of shell access in environment requirements",
          "source": "builtin"
        }
      ],
      "tools_used": ["builtin"],
      "details": "CRITICAL: Command injection vulnerability. User SQL input passed directly to psql shell command without sanitization.",
      "metadata": {
        "pass": false,
        "is_gate": true,
        "checks_executed": ["hardcoded_secrets", "external_calls", "privilege_escalation", "command_injection", "prompt_injection", "data_exfiltration", "excessive_permissions"],
        "tools_used": ["builtin"]
      }
    },
    "functional": {
      "score": 4,
      "max": 10,
      "tier": "A",
      "details": "Basic query optimization flow defined but lacks edge case handling, no output format specified, error handling minimal.",
      "sub_scores": {
        "core_functionality": 5,
        "edge_handling": 3,
        "output_stability": 4,
        "output_quality": 4,
        "error_handling": 2,
        "instruction_clarity": 5
      },
      "issues": [
        {
          "category": "edge_handling",
          "severity": "warning",
          "item": "No handling for invalid SQL input",
          "location": "instructions"
        },
        {
          "category": "output_stability",
          "severity": "warning",
          "item": "No output format defined — optimization suggestions vary in structure",
          "location": "instructions"
        },
        {
          "category": "error_handling",
          "severity": "error",
          "item": "No error handling for database connection failures",
          "location": "instructions"
        }
      ],
      "metadata": {
        "tier": "A",
        "skill_type": "atom",
        "test_cases": [
          "Simple SELECT optimization → should suggest index",
          "Complex JOIN query → should analyze query plan",
          "Invalid SQL → unknown behavior",
          "Empty input → unknown behavior"
        ]
      }
    },
    "comparative": {
      "score": 3,
      "max": 10,
      "details": "Moderate improvement for complex queries where tool integration helps, but minimal advantage for simple optimizations that LLMs handle well.",
      "sub_scores": {},
      "issues": [],
      "metadata": {
        "delta": 0.12,
        "value_category": "domain_expertise",
        "scenarios": [
          {
            "name": "Simple index suggestion",
            "with_skill": 6,
            "without_skill": 5,
            "improvement": 0.1
          },
          {
            "name": "Multi-table JOIN optimization",
            "with_skill": 5,
            "without_skill": 4,
            "improvement": 0.1
          },
          {
            "name": "Query plan analysis with EXPLAIN",
            "with_skill": 6,
            "without_skill": 4,
            "improvement": 0.2
          }
        ]
      }
    },
    "uniqueness": {
      "score": 7,
      "max": 10,
      "details": "Good uniqueness. PostgreSQL-specific optimization with potential tool integration (EXPLAIN ANALYZE). No known overlapping skills in registry.",
      "similar_skills": [],
      "supersession_risk": "low",
      "sub_scores": {},
      "issues": [],
      "metadata": {
        "supersession_risk": "low",
        "modifiers_applied": ["+1 tool-integration potential"],
        "registry_available": false
      }
    }
  },
  "overall_score": 38,
  "verdict": "FAIL",
  "weakest_dimension": "security",
  "recommendations": [
    "CRITICAL: Fix command injection vulnerability in SQL execution (D3)",
    "Add not_for boundaries to description to reduce false triggers (D2)",
    "Define output format for optimization suggestions (D4)",
    "Add error handling for database connection failures (D4)"
  ],
  "metadata": {
    "evaluated_at": "2026-01-15T10:30:00Z",
    "evaluator_version": "1.0.0",
    "feedback_applied": false
  }
}
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall