name: evidence-collector description: Standardized evidence collection skill for PM review agents. Provides consistent patterns for gathering, formatting, and verifying evidence across all categories.

Evidence Collector Skill

This skill provides standardized evidence collection patterns for all PM review agents. Ensures consistent, verifiable evidence that supports improvement recommendations.

Evidence Collection Protocol

Phase 1: Discovery Evidence

When discovering an issue, collect initial evidence:

json

{
  "discovery_evidence": {
    "source": "grep|glob|read|manual",
    "search_pattern": "<pattern-used>",
    "files_searched": ["<list>"],
    "matches_found": [
      {
        "file": "src/api/route.ts",
        "line": 45,
        "content": "<matched-content>",
        "context": "<surrounding-lines>"
      }
    ],
    "discovery_timestamp": "<ISO-8601>"
  }
}

Phase 2: Verification Evidence

After discovery, verify the issue exists:

json

{
  "verification_evidence": {
    "verification_method": "file_read|git_check|pattern_match|ast_analysis",
    "verified_at": "<ISO-8601>",
    "verification_result": "confirmed|refuted|inconclusive",
    "verification_details": {
      "file_exists": true,
      "pattern_confirmed": true,
      "git_tracked": true,
      "not_in_gitignore": true
    }
  }
}

Phase 3: Impact Evidence

Quantify the impact of the issue:

json

{
  "impact_evidence": {
    "affected_files": ["<list>"],
    "affected_functions": ["<list>"],
    "usage_count": N,
    "downstream_impact": ["<components-affected>"],
    "user_facing": true|false,
    "data_risk": "none|low|medium|high|critical"
  }
}

Category-Specific Evidence Templates

Security Evidence

json

{
  "security_evidence": {
    "vulnerability_class": "A01_broken_access_control",
    "owasp_category": "A01:2021 - Broken Access Control",
    "cwe_id": "CWE-639",
    "attack_vector": "<description-of-exploitation>",
    "severity": "critical|high|medium|low",
    "exploitability": {
      "requires_auth": true|false,
      "network_accessible": true|false,
      "user_interaction": true|false
    },
    "git_verification": {
      "file_tracked": true,
      "not_ignored": true,
      "command_used": "git ls-files <file>"
    },
    "remediation": {
      "fix_location": "src/api/route.ts:45",
      "fix_code": "<suggested-fix>",
      "effort_estimate": "hours|days|weeks"
    }
  }
}

Performance Evidence

json

{
  "performance_evidence": {
    "issue_type": "n_plus_one|missing_index|bundle_size|render_blocking",
    "measurement_method": "<how-to-measure>",
    "baseline_metric": {
      "value": N,
      "unit": "ms|KB|count|requests"
    },
    "expected_improvement": {
      "value": N,
      "unit": "ms|KB|count|requests",
      "percentage": "X%"
    },
    "code_pattern": {
      "file": "src/lib/db.ts",
      "line": 23,
      "pattern": "N+1 query in loop",
      "fix_pattern": "Use batch query with IN clause"
    }
  }
}

UX Evidence

json

{
  "ux_evidence": {
    "journey_traced": true,
    "friction_point": {
      "step": "checkout → payment",
      "issue": "No loading feedback during payment processing",
      "user_impact": "Users click submit multiple times"
    },
    "friction_score": 7,
    "cognitive_load": {
      "decisions_required": 3,
      "fields_to_complete": 8,
      "navigation_steps": 4
    },
    "accessibility": {
      "wcag_violations": ["missing-alt-text", "low-contrast"],
      "keyboard_navigable": false,
      "screen_reader_friendly": false
    }
  }
}

UI Evidence

json

{
  "ui_evidence": {
    "design_system_check": {
      "uses_design_tokens": false,
      "hardcoded_values": ["#333333", "16px"],
      "inconsistent_patterns": ["button-styles", "spacing"]
    },
    "visual_issues": [
      {
        "component": "Button",
        "file": "src/components/Button.tsx",
        "issue": "Hardcoded color instead of theme token",
        "fix": "Replace #333 with var(--color-text-primary)"
      }
    ],
    "responsive_check": {
      "breakpoints_tested": ["mobile", "tablet", "desktop"],
      "issues": ["Overflow on mobile at 320px"]
    }
  }
}

API Evidence

json

{
  "api_evidence": {
    "endpoint": "POST /api/users",
    "issues": {
      "missing_auth": true,
      "missing_validation": true,
      "missing_rate_limit": true,
      "missing_pagination": false
    },
    "request_analysis": {
      "params_validated": false,
      "body_validated": false,
      "dangerous_operations": ["DELETE without auth"]
    },
    "response_analysis": {
      "error_format_consistent": false,
      "data_exposure_risk": ["password_hash returned"]
    }
  }
}

Data Evidence

json

{
  "data_evidence": {
    "schema_issues": {
      "missing_indexes": ["user_id on orders"],
      "missing_constraints": ["foreign_key on user_id"],
      "type_mismatches": []
    },
    "query_analysis": {
      "n_plus_one_patterns": ["getUser in loop"],
      "missing_limits": ["SELECT * without LIMIT"],
      "unoptimized_joins": []
    },
    "migration_safety": {
      "data_loss_risk": false,
      "downtime_required": false,
      "rollback_plan": "<description>"
    },
    "rls_audit": {
      "tables_checked": ["users", "orders"],
      "rls_enabled": { "users": true, "orders": false },
      "policies_verified": { "users": true, "orders": false }
    }
  }
}

Code Quality Evidence

json

{
  "code_quality_evidence": {
    "complexity_metrics": {
      "cyclomatic_complexity": 15,
      "cognitive_complexity": 23,
      "lines_of_code": 450,
      "nesting_depth": 5
    },
    "code_smells": [
      {
        "type": "long_method",
        "file": "src/lib/processor.ts",
        "method": "processData",
        "lines": 180
      }
    ],
    "duplication": {
      "duplicated_blocks": 3,
      "files_affected": ["a.ts", "b.ts", "c.ts"],
      "suggested_abstraction": "Extract to shared utility"
    },
    "tech_debt": {
      "todo_count": 12,
      "fixme_count": 5,
      "hack_count": 2
    }
  }
}

Evidence Collection Commands

Quick Evidence Collection (Tier A)

bash

# Find all occurrences of a pattern
grep -rn "pattern" --include="*.ts" --include="*.tsx" .

# Get file context (5 lines before/after)
grep -B5 -A5 "pattern" <file>

# Check git tracking
git ls-files <file>
git check-ignore -q <file>

# Find related files
find . -name "*related*" -type f

# Count occurrences
grep -c "pattern" <file>

Deep Evidence Collection (Tier B - Explore Agent)

code

Invoke Explore agent for:
- Import/dependency graph analysis
- Cross-file pattern detection
- Architectural impact assessment
- Test coverage analysis

Evidence Validation Rules

Required Evidence by Category

Category	Required Evidence Fields
Security	git_verification, attack_vector, owasp_category
Performance	measurement_method, baseline_metric, expected_improvement
UX	journey_traced, friction_score, user_impact
UI	design_system_check, visual_issues
API	endpoint, issues object, request_analysis
Data	schema_issues OR query_analysis, rls_audit if applicable
Code Quality	complexity_metrics, code_smells
Feature	user_value, technical_feasibility

Evidence Quality Score

Calculate evidence quality (0.0 - 1.0):

code

evidence_quality = (
  file_references_exist * 0.3 +
  line_numbers_provided * 0.2 +
  fix_location_specified * 0.2 +
  impact_quantified * 0.15 +
  verification_completed * 0.15
)

Reject findings with evidence_quality < 0.6

Evidence Output Format

Final evidence structure for each finding:

json

{
  "evidence": {
    "confidence_score": 0.85,
    "evidence_quality": 0.78,
    "verification_tier": "tier2",
    "category_evidence": {
      /* category-specific */
    },
    "discovery_evidence": {
      /* how found */
    },
    "verification_evidence": {
      /* how verified */
    },
    "impact_evidence": {
      /* scope of impact */
    }
  }
}

Usage Instructions

•Discovery: Use grep/glob to find potential issues
•Verification: Run verification checks (git, context, mitigations)
•Impact: Trace affected files and downstream components
•Format: Structure evidence using category-specific template
•Validate: Check evidence quality score >= 0.6
•Include: Attach evidence to finding before submission