Config

Config — Environment & Configuration Management

Activate When

• User invokes /godmode:config
• User says "manage environments", "config validation", "feature flags"
• User needs dev/staging/prod parity checking
• User wants to design a feature flag system or A/B test rollout
• Ship skill needs environment verification before deployment
• User asks "are my environments in sync?" or "check config drift"

Workflow

Step 1: Inventory Current Configuration

Map all configuration sources and environments:

# Find config files
find . -name "*.env*" -o -name "*.config.*" -o -name "*.yml" -o -name "*.yaml" -o -name "*.toml" -o -name
"*.ini" | grep -v node_modules | grep -v .git

# Check for environment-specific files
find . -name "*development*" -o -name "*staging*" -o -name "*production*" -o -name "*prod*" -o -name "*dev*" |
grep -v node_modules | grep -v .git

CONFIG INVENTORY:
Environments: <dev | staging | prod | custom>
Config sources:
  - Environment variables: <list of .env files>
  - Config files: <list of config files>
  - Secret managers: <vault/AWS SSM/GCP Secret Manager/none>
  - Feature flag provider: <LaunchDarkly/Unleash/custom/none>
Config format: <JSON | YAML | TOML | dotenv | mixed>
Total config keys: <N>
Secret keys: <N>
Non-secret keys: <N>

Step 2: Environment Parity Check

Compare configurations across environments to detect drift:

Key-Level Comparison

PARITY CHECK:
| Config Key | Dev | Staging | Prod |
|--|--|--|--|
| DATABASE_URL | ✓ | ✓ | ✓ |
| REDIS_URL | ✓ | ✓ | ✓ |
│ LOG_LEVEL           │ debug│ info    │ warn │  ← EXPECTED DIFF
│ FEATURE_NEW_UI      │ true │ true    │ false│  ← EXPECTED DIFF
│ MAX_CONNECTIONS     │ 10   │ 50      │ 100  │  ← EXPECTED DIFF
| API_TIMEOUT_MS | 5000 | 5000 | 5000 |
│ SENTRY_DSN          │ ✓    │ ✓       │ ✗    │  ← MISSING IN PROD
│ NEW_SERVICE_URL     │ ✓    │ ✗       │ ✗    │  ← ONLY IN DEV

Drift Categories

CRITICAL DRIFT (must fix):
- Keys present in one env but missing in another (likely deployment failure)
- Type mismatches (string in dev, number in prod)
- Secret keys with placeholder values in non-dev environments

EXPECTED DRIFT (document and accept):
- Log levels (debug in dev, warn in prod)
- Connection pool sizes (scaled per environment)
- Feature flags (intentional per-environment rollout)
- Debug/profiling settings (dev-only)

SUSPICIOUS DRIFT (investigate):
- Different values for same key with no documented reason
- Timeout or retry values that differ without scaling justification
- Third-party service URLs that don't match environment tier

Step 3: Config Validation Schema

Generate or verify a validation schema for all configuration:

// config/schema.ts — Single source of truth for all config keys
const configSchema = {
  DATABASE_URL: {
    type: 'string',
    required: true,
    format: 'uri',

Validation Rules

For EVERY config key, validate:
1. PRESENCE — Required keys exist in every environment
2. TYPE — Value matches expected type (string, number, boolean, URL, etc.)
3. FORMAT — Value matches pattern (URLs, email, API key formats)
4. RANGE — Numeric values within acceptable bounds
5. SENSITIVITY — Sensitive values not hardcoded or committed to git
6. CONSISTENCY — Same key has same type across all environments

Startup Validation

// Validate config on application startup — fail fast
function validateConfig(env: Record<string, string>): void {
  const errors: string[] = [];
  for (const [key, schema] of Object.entries(configSchema)) {
    const value = env[key];
    if (schema.required && !value) {

IF config change breaks health check: rollback immediately. WHEN feature flag stale >30 days: schedule removal.

Step 4: Feature Flag Design

Design and manage feature flags for controlled rollouts:

Flag Types

FLAG TYPES:
1. RELEASE FLAG — Gate new features (temporary, remove after full rollout)
   Example: FEATURE_NEW_CHECKOUT=true
   Lifecycle: Create → Dev → Staging → % Prod → 100% Prod → Remove flag

2. EXPERIMENT FLAG — A/B test with measurement
   Example: EXPERIMENT_PRICING_V2={variant: "B", percentage: 25}
   Lifecycle: Create → Configure variants → Run → Measure → Pick winner → Remove

3. OPS FLAG — Control operational behavior
   Example: OPS_MAINTENANCE_MODE=false
   Lifecycle: Create → Toggle during incidents → Keep permanently

4. PERMISSION FLAG — Gate features by user segment
   Example: PERMISSION_BETA_FEATURES=["user_123", "org_456"]

Flag Schema

interface FeatureFlag {
  name: string;                          // SCREAMING_SNAKE_CASE
  type: 'release' | 'experiment' | 'ops' | 'permission';
  description: string;                   // What this flag controls
  owner: string;                         // Team or person responsible
  createdAt: string;                     // ISO date

Flag Lifecycle Management

Every flag: owner + expiry date. Release flags >30 days at 100%: remove flag, keep code. Experiment flags >14 days: conclude, pick winner. Dead flags (no code refs): delete. Keep total under 20 for small teams. Weekly stale flag report.

Step 5: A/B Test Setup

Design controlled experiments with statistical rigor:

Experiment Design

EXPERIMENT PLAN:
Name: <experiment_name>
Hypothesis: "Changing <X> will improve <metric> by <expected_delta>"
Primary metric: <conversion_rate | revenue | engagement | latency | etc.>
Secondary metrics: <list of guardrail metrics to monitor>
Minimum detectable effect: <smallest meaningful improvement, e.g., 2%>
Statistical significance: <p-value threshold, default 0.05>
Required sample size: <calculated from MDE and baseline>

Variants:
  Control (A): <current behavior>
  Treatment (B): <new behavior>
  [Treatment (C)]: <optional additional variant>

Traffic split: <50/50 | 80/20 | custom>

Rollout Strategy

Phase 1: Internal (100%, 2-3 days, catch bugs). Phase 2: Canary (1-5%, 24-48h, verify no regressions). Phase 3: Controlled (10% -> 25% -> 50%, 1-2 weeks per increment, gather statistical significance).

Step 6: Secret Management Audit

Verify secrets are handled safely across all environments:

SECRET AUDIT:
| Check | Status | Finding |
|--|--|--|
| .env in .gitignore | PASS/FAIL | <detail> |
| No secrets in code | PASS/FAIL | <files with hardcoded> |
| No secrets in logs | PASS/FAIL | <log statements to fix> |
| Secrets rotatable | PASS/FAIL | <non-rotatable secrets> |
| Secrets have expiry | PASS/FAIL | <non-expiring secrets> |
| Dev ≠ prod secrets | PASS/FAIL | <shared secrets> |
| Secret manager in use | PASS/FAIL | <recommendation> |
| Encryption at rest | PASS/FAIL | <unencrypted stores> |

Step 7: Generate Config Report

  CONFIG AUDIT — <project>
  Environments: <N> configured
  Total config keys: <N>
  Sensitive keys: <N>
  PARITY:
  Keys in all envs: <N>/<total>
  Missing keys: <N> (CRITICAL)
  Expected drift: <N> (documented)
  Suspicious drift: <N> (needs investigation)
  VALIDATION:
  Schema coverage: <X>% of keys have validation

Step 8: Commit and Transition

1. Save report as docs/config/<project>-config-audit.md
2. Save validation schema if generated
3. Commit: "config: <project> — <verdict> (<N> keys, <N> flags, <N> issues)"
4. If CRITICAL: "Missing keys in production or secrets exposed. Fix immediately."
5. If HEALTHY: "Configuration is consistent. Ready for deployment."

Key Behaviors

1. Never commit secrets. Flag as CRITICAL immediately.
2. Schema is source of truth. Type, validation, description.
3. Parity before deploy. Fail fast on missing keys.
4. Flags have lifecycles. Owner + expiry date required.
5. A/B tests need math. Sample size before launch.
6. Environment drift is a bug. Document or fix.

Flags & Options

HARD RULES

Never ask to continue. Loop autonomously until all environments are audited and drift is resolved.

1. NEVER commit secrets to source control. If found, flag as CRITICAL immediately.
2. NEVER deploy to an environment with missing required config keys. Fail fast.
3. EVERY config key MUST have a schema entry with type, validation, and description.
4. EVERY feature flag MUST have an owner and expiry date.
5. NEVER add a flag without a cleanup plan — document removal conditions at creation time.
6. git commit BEFORE verify — commit config changes, then validate against schema.
7. Automatic revert on regression — if config change causes startup failure, revert immediately.
8. TSV logging — log every config audit:

   timestamp    environments    total_keys    missing_keys    secret_issues    flags_stale    verdict
   

Auto-Detection

On activation, automatically detect all configuration without asking:

AUTO-DETECT:
1. Config files:
   find . -name "*.env*" -o -name "*.config.*" -o -name "*.yml" \
     -o -name "*.yaml" -o -name "*.toml" -o -name "*.ini" \
     | grep -v node_modules | grep -v .git

2. Environment-specific files:
   find . -name "*development*" -o -name "*staging*" -o -name "*production*" \
     -o -name "*prod*" -o -name "*dev*" | grep -v node_modules

3. Secret references:
   grep -r "SECRET\|API_KEY\|PASSWORD\|TOKEN\|PRIVATE" \
     --include="*.env*" --include="*.config.*" -l

4. Feature flag provider:

Output Format

Print on completion: Config: {config_key_count} keys across {env_count} environments. Secrets: {secret_count} (all in secret manager: {secret_mgr_status}). Drift: {drift_count} keys differ. Validation: {validation_status}. Feature flags: {flag_count}. Verdict: {verdict}.

TSV Logging

Log every configuration operation to .godmode/config-results.tsv:

iteration	task	environment	keys_total	secrets_count	drift_detected	validation_status	status
1	inventory	production	45	12	0	passing	audited
2	inventory	staging	45	12	3	passing	drift_found
3	secrets	all	0	12	0	migrated	migrated
4	validation	all	45	0	0	zod_schema	configured

Columns: iteration, task, environment, keys_total, secrets_count, drift_detected, validation_status, status(audited/drift_found/migrated/configured/failed).

Success Criteria

All keys inventoried. Secrets in secret manager. Startup validation fails fast. Drift detected and explained. Flags have expiry + cleanup plans. Typed config parsing (no raw process.env). Config changes auditable.

Error Recovery

Keep/Discard

KEEP if: improvement verified. DISCARD if: regression or no change. Revert discards immediately.

Stop Conditions

Stop when: target reached, budget exhausted, or >5 consecutive discards.