Configuring Code Analyzer Skill

Overview

Ecosystem: This skill is part of a 3-skill Code Analyzer suite — dx-code-analyzer-run (scans & results) · dx-code-analyzer-configure (setup, config, CI/CD) · dx-code-analyzer-custom-rule-create (custom rule authoring).

This skill manages the code-analyzer.yml configuration file — the single source of truth for how Code Analyzer behaves in a project. All customization (engines, rules, ignores, suppressions) is done by creating or editing this file. If the file doesn't exist, this skill creates it in the current working directory.

Scope

In scope:

• Checking prerequisites (sf CLI, Java, Node.js, Python, org auth)
• Installing/updating the Code Analyzer plugin
• Creating code-analyzer.yml if it doesn't exist
• Editing code-analyzer.yml for all configuration changes
• Engine settings, rule overrides, ignore patterns, suppressions
• CI/CD pipeline setup (GitHub Actions, Jenkins, etc.)
• Environment validation and troubleshooting

Out of scope:

• Running scans → use dx-code-analyzer-run skill
• Fixing violations, explaining rules, suppression management → use dx-code-analyzer-run skill
• Creating custom rules → use dx-code-analyzer-custom-rule-create skill

Tool Usage Rules

Allowed: Bash (sf, java, node, python3, git, npm), Read, Write, Edit Forbidden: MCP tools, Agent tool, Web tools, other skills, which, find, locate, searching for binaries

Core Principle: YAML Only When Customizing

Code Analyzer works out of the box with NO config file — all defaults are built into the tool. The code-analyzer.yml file is ONLY created when the user explicitly requests a customization.

Rules:

• Do NOT create code-analyzer.yml proactively — only when user asks to change something
• Do NOT duplicate built-in defaults — only write entries that intentionally override behavior
• Always place at project root — where sfdx-project.json or sf-project.json lives
• The CLI auto-discovers it — sf code-analyzer run from project root automatically picks up code-analyzer.yml in that directory. No --config-file flag needed.
• User says "configure code analyzer" with no specifics? → Ask what they want to customize. Don't create an empty or boilerplate file.

Workflow:

1. User requests a customization (e.g., "disable PMD", "ignore test files", "increase SFGE memory")
2. Check if code-analyzer.yml exists at project root
3. If NO → create it at project root with ONLY the requested override
4. If YES → read it, then edit in the requested change
5. Validate with sf code-analyzer config

Step 1: Understand Intent and Map to Config Sections

The user can request ANY combination of configuration changes in natural language. Your job is to:

1. Parse what they want — may be one thing or many things combined
2. Map each request to the correct section(s) of code-analyzer.yml
3. Create the file if it doesn't exist, then apply all changes

The code-analyzer.yml Structure (what you can write/edit)

config_root: .                    # Root for relative path resolution
log_folder: <path>                # Where logs are written
log_level: <1-5>                  # 1=Error, 2=Warn, 3=Info, 4=Debug, 5=Fine

ignores:                          # Files/folders excluded from scanning
  files: [<glob patterns>]

engines:                          # Per-engine settings
  <engine_name>:
    disable_engine: <bool>
    <engine_specific_keys>: ...

rules:                            # Per-rule overrides
  <engine_name>:
    <rule_name>:
      severity: <1-5>
      tags: [<strings>]
      disabled: <bool>

suppressions:                     # Bulk suppression configuration
  disable_suppressions: <bool>
  "<file_or_folder_path>":
    - rule_selector: "<selector>"
      max_suppressed_violations: <number|null>
      reason: "<why>"

Mapping Principle

Any user request maps to one or more sections above. Parse the intent and edit the right section(s):

File Existence Decision

BEFORE editing anything, check if code-analyzer.yml exists at project root:

ls code-analyzer.yml code-analyzer.yaml 2>/dev/null

• File does NOT exist → Create it at project root with ONLY the user's requested override(s)
• File exists → Read it, then Edit to add/modify the requested section(s)

The CLI auto-discovers code-analyzer.yml in the current directory. Since scans run from project root, the file must live there.

Rule Name Resolution — ALWAYS Before Writing YAML

When a user references rules by partial, descriptive, or approximate names (e.g., "the doc rule", "CRUD violation", "console rule", "hardcoded values"), you MUST resolve to exact rule names using the lookup in Step 6.1 BEFORE writing any YAML. The code-analyzer.yml file silently ignores rule names that don't exactly match — there is no error, the override just won't apply.

Examples of fuzzy → exact resolution needed:

• "Disable the ApexDoc rule" → lookup confirms ApexDoc (engine: pmd)
• "Demote no-console to low" → lookup confirms no-console (engine: eslint)
• "Make CRUD violations critical" → lookup confirms ApexCRUDViolation (engine: pmd)
• "Turn off the hardcoded values check" → lookup finds @salesforce-ux/slds/no-hardcoded-values-slds2 (engine: eslint)
• "Disable the injection rule" → multiple matches possible → ask user which one

Only skip the lookup when the user provides an unambiguous, exact, well-known name (e.g., "ApexDoc", "no-console", "no-unused-vars").

Handling Combined/Complex Requests

Users will often combine multiple changes in one request. Handle ALL of them in a single edit:

• "Disable PMD's ApexDoc rule and make CRUD violations critical" → edit two entries under rules.pmd
• "Exclude test files and vendor code, and increase SFGE memory" → edit ignores.files + engines.sfge.java_max_heap_size
• "Set up code analyzer with only ESLint and PMD, ignore node_modules" → create file with engines (disable others) + ignores
• "Make all security rules severity 1" → look up rules via sf code-analyzer rules --rule-selector Security, then override each
• "Configure code analyzer" (no specifics) → ask user what they want to customize before creating any file

Quick Reference: Common Requests → Config Output

The AI must understand the YAML schema and write valid config for ANY request, not just the examples above.

Step 2: Check Prerequisites and Install

Run bash "<skill_dir>/scripts/check-prerequisites.sh" or check manually:

sf --version 2>&1                                    # sf CLI
sf plugins --core 2>&1 | grep -i "code-analyzer"    # Plugin
java -version 2>&1                                   # Java 11+ (PMD, CPD, SFGE)
node --version 2>&1                                  # Node 18+ (ESLint, RetireJS)
python3 --version 2>&1                               # Python 3 (Flow engine)

If anything is missing, install it (always ask user first):

npm install -g @salesforce/cli                       # sf CLI
sf plugins install @salesforce/plugin-code-analyzer  # Code Analyzer plugin

For Java/Node/Python installs, read <skill_dir>/references/engine-prerequisites.md. If install fails, read <skill_dir>/references/troubleshooting.md.

Step 2A: Diagnose and Fix a Broken Setup

TRIGGER: User says "not working", "broken", "getting errors", "scan fails", "help me fix", etc.

Read <skill_dir>/references/diagnostic-flow.md for the complete layered diagnostic procedure, fix table, and anti-patterns.

Key principles (always apply):

• Never search for binaries (which, find, ls /opt/homebrew/bin/)
• Never use sfdx as a workaround — only sf
• Fix layer by layer: CLI → Plugin → Engine deps → verify scan
• Give user ONE command at a time, wait for confirmation before continuing
• After fix succeeds, proceed to run the full scan automatically

Step 3: Create or Edit code-analyzer.yml

Only triggered when user requests a customization. Never create proactively.

Creating (file doesn't exist)

Choose one of the two approaches below — do not run both:

Option A — Auto-generate from project type (recommended for first-time setup):

Run bash "<skill_dir>/scripts/generate-config.sh". This detects Apex, LWC, and Flow markers and produces a minimal code-analyzer.yml suited to the project. Skip to the "After any create/edit, validate" section.

Note: The script exits with an error if code-analyzer.yml already exists. Delete the existing file first if you need to regenerate.

Option B — Write manually (when the user has specific customizations in mind):

Read the appropriate example config as a reference for structure:

• For Apex-only projects, read <skill_dir>/examples/apex-project-config.yml
• For LWC-only projects, read <skill_dir>/examples/lwc-project-config.yml
• For full-stack (Apex + LWC + Flows), read <skill_dir>/examples/fullstack-project-config.yml

Write the file at project root using the Write tool. Include ONLY the user's requested changes:

# Example: user said "ignore test files and increase SFGE memory"
# → Write to project root (where sfdx-project.json lives):

ignores:
  files:
    - "**/test/**"
    - "**/__tests__/**"

engines:
  sfge:
    java_max_heap_size: "4g"

Do NOT add config_root, log_folder, or any other field the user didn't ask for.

Editing (file already exists)

Read the file, then use the Edit tool to add/modify only the relevant section. Preserve everything else.

After any create/edit, validate:

Run bash "<skill_dir>/scripts/validate-config.sh" to validate YAML syntax and schema correctness, or use the CLI directly:

sf code-analyzer config

(No --config-file needed — the CLI auto-discovers code-analyzer.yml in CWD.)

If user says "configure code analyzer" with no specifics

Ask: "What would you like to customize? For example: ignore certain files, change rule severities, tune engine settings, or disable engines you don't need."

Step 4: Enable/Disable Engines

Edit the engines section in code-analyzer.yml:

engines:
  pmd:
    disable_engine: true       # Disable PMD
  eslint:
    disable_engine: false      # Enable ESLint (default)

Valid engine names: pmd, cpd, eslint, regex, retire-js, flow, sfge, apexguru

Always validate after editing:

sf code-analyzer config --config-file code-analyzer.yml

Step 5: Ignore Patterns

Edit the ignores section in code-analyzer.yml:

ignores:
  files:
    - "**/node_modules/**"
    - "**/.sfdx/**"
    - "**/.sf/**"
    - "**/vendor/**"
    - "**/*.min.js"

Common patterns:

Step 6: Rule Overrides

Edit the rules section in code-analyzer.yml. Each rule can have severity, tags, and disabled overrides:

rules:
  pmd:
    ApexCRUDViolation:
      severity: 1              # Promote to Critical
    AvoidGlobalModifier:
      disabled: true           # Turn off entirely
    ApexDoc:
      severity: 5              # Demote to Info
      tags: ["Documentation"]
  eslint:
    no-console:
      severity: 4              # Demote to Low
    no-unused-vars:
      severity: 2              # Promote to High

Severity values: 1/Critical, 2/High, 3/Moderate, 4/Low, 5/Info

6.1 Rule Name Resolution (Fuzzy Matching)

⚠️ CRITICAL: A misspelled or partial rule name in code-analyzer.yml is SILENTLY IGNORED — no error, the override just won't apply.

When users reference rules by approximate names (e.g., "the doc rule", "CRUD violation", "hardcoded values"), resolve to exact names BEFORE writing YAML:

sf code-analyzer rules --rule-selector all 2>&1 | grep -i "<USER_KEYWORD>"

• 1 match → use that exact name + its engine for the YAML path
• Multiple matches → ask user which one they meant
• 0 matches → try broader keywords or inform user

Skip the lookup only when the name is unambiguous and exact (e.g., "ApexDoc", "no-console", "no-unused-vars").

For detailed matching strategies, common fuzzy→exact mappings, and engine identification: Read <skill_dir>/references/rule-name-resolution.md.

Step 7: Engine-Specific Settings

Edit the engines section. Most common overrides:

engines:
  sfge:
    java_max_heap_size: "4g"      # <200 classes→"2g", 200-500→"4g", 500+→"6g"/"8g"
    java_thread_count: 4
    java_thread_timeout: 900000
  eslint:
    auto_discover_eslint_config: true    # Use project's own ESLint config
    eslint_config_file: "./eslint.config.mjs"
  pmd:
    custom_rulesets: ["./config/custom-pmd-rules.xml"]
    java_classpath_entries: ["./lib/custom-rules.jar"]
  cpd:
    minimum_tokens: { apex: 100, javascript: 100 }
  apexguru:
    target_org: "my-org-alias"
  flow:
    python_command: "python3"
  # regex.custom_rules — use the dx-code-analyzer-custom-rule-create skill to create these.
  # Never hand-write regex patterns into code-analyzer.yml: quotes/backslashes
  # inside YAML cause parsing failures. The create-regex-rule.js script handles
  # serialization correctly and must always be used for regex rule creation.

For full property list per engine, read <skill_dir>/references/config-schema.md.

Step 8: CI/CD Pipeline Setup

Detect CI system from workspace (.github/workflows/ → GitHub Actions, Jenkinsfile → Jenkins, etc.). Read <skill_dir>/references/ci-cd-templates.md for templates. Use <skill_dir>/examples/ci-github-actions.yml as GitHub Actions base. Key flags: --severity-threshold 2 (gate), --output-file results.sarif (GitHub scanning), --config-file code-analyzer.yml.

Step 9: View Current Configuration

sf code-analyzer config                               # Show effective config
sf code-analyzer config --rule-selector pmd:Security  # Specific rules
sf code-analyzer config --include-unmodified-rules    # All defaults

Cross-Skill Integration

This skill works together with dx-code-analyzer-run. The AI agent should seamlessly hand off between them:

When dx-code-analyzer-run delegates HERE:

If a user says "scan my code" / "run code analyzer" but it fails (CLI missing, plugin not installed, or scan errors out), dx-code-analyzer-run delegates to this skill. In that case:

1. Run the diagnose and fix flow (Step 2A) — find what's broken, fix it
2. After everything works, automatically proceed to run the scan — do not stop and ask. The user's original intent was to scan.
3. Hand execution back to dx-code-analyzer-run behavior (build command, execute, parse results).

When THIS skill hands off to dx-code-analyzer-run:

After any successful configuration action, offer to run a scan (e.g., "Setup complete! Want me to run a scan?", "Config updated — want to scan and verify?"). If user says yes, proceed with dx-code-analyzer-run behavior.

When THIS skill hands off to dx-code-analyzer-custom-rule-create:

If the user asks to create a custom rule while you are working on configuration (e.g., "also add a rule that bans System.debug", "set up a PMD XPath rule"), delegate to dx-code-analyzer-custom-rule-create. When that skill finishes, the custom_rulesets or eslint_config_file pointer in code-analyzer.yml may need to be set — that edit belongs here, in this skill.

When user intent spans BOTH skills:

Handle end-to-end: "not working" → Diagnose → Fix → Scan. "Set up and scan" → Install → Scan. "Disable ESLint and scan Apex" → Edit config → Run with --rule-selector pmd. "Configure custom rules and scan" → dx-code-analyzer-custom-rule-create → wire config → dx-code-analyzer-run. Always follow through to the user's final intent.

Rules / Constraints

Gotchas

For full troubleshooting, read <skill_dir>/references/troubleshooting.md.

Reference File Index

<skill_dir> is the absolute path to the directory containing this SKILL.md file.