Cloud Custodian Management Skill

Manage and inspect Cloud Custodian policies, resource filters, actions, and compliance reports.

MANDATORY: Discovery-First Pattern

Always run policies in dry-run mode before enforcing actions.

Phase 1: Discovery

#!/bin/bash
export AWS_PAGER=""

echo "=== Custodian Version ==="
custodian version 2>/dev/null

echo ""
echo "=== Policy Files ==="
find . -name "*.yml" -o -name "*.yaml" 2>/dev/null | while read f; do
    POLICIES=$(grep -c "^  - name:" "$f" 2>/dev/null || grep -c "^- name:" "$f" 2>/dev/null)
    [ "$POLICIES" -gt 0 ] && echo "$f: $POLICIES policies"
done | head -15

echo ""
echo "=== Policy Summary ==="
for f in $(find . -name "*.yml" -o -name "*.yaml" 2>/dev/null | head -5); do
    grep -E "^\s+- name:|^\s+resource:" "$f" 2>/dev/null | paste - - | head -10
done | head -20

echo ""
echo "=== Output Directory ==="
ls -la output/ 2>/dev/null | head -10 || echo "No output directory found"

Phase 2: Analysis

#!/bin/bash
export AWS_PAGER=""

POLICY_FILE="${1:?Policy file required}"

echo "=== Validate Policies ==="
custodian validate "$POLICY_FILE" 2>&1 | tail -10

echo ""
echo "=== Dry Run ==="
custodian run --dryrun -s output/ "$POLICY_FILE" 2>&1 | tail -20

echo ""
echo "=== Resources Found ==="
for dir in output/*/; do
    POLICY_NAME=$(basename "$dir")
    COUNT=$(cat "$dir/resources.json" 2>/dev/null | jq 'length' 2>/dev/null || echo "0")
    echo "$POLICY_NAME: $COUNT resources matched"
done | head -15

echo ""
echo "=== Sample Matches ==="
for dir in output/*/; do
    POLICY_NAME=$(basename "$dir")
    echo "--- $POLICY_NAME ---"
    cat "$dir/resources.json" 2>/dev/null | jq '.[0:3][] | {id: (.ResourceId // .InstanceId // .BucketName // .Name // .id), type: (.ResourceType // .InstanceType // "N/A")}' 2>/dev/null | head -10
done | head -25

Output Format

CLOUD CUSTODIAN STATUS: <policy-file>
Policies: <count> | Resource Types: <unique types>
Validation: <passed|failed>
Dry Run Results:
  <policy-name>: <count> resources matched
  <policy-name>: <count> resources matched
Actions: <list of actions configured>
Issues: <any validation errors, empty matches, or action warnings>

Safety Rules

• NEVER run policies without --dryrun first -- always preview matched resources
• NEVER use destructive actions (terminate, delete) without explicit user confirmation
• Always validate policy files before running -- custodian validate
• Review matched resources from dry-run output before enabling enforcement
• Test policies on non-production accounts before deploying to production

Anti-Hallucination Rules

1. NEVER assume resource names — always discover via CLI/API in Phase 1 before referencing in Phase 2.
2. NEVER fabricate metric names or dimensions — verify against the service documentation or --help output.
3. NEVER mix CLI commands between service versions — confirm which version/API you are targeting.
4. ALWAYS use the discovery → verify → analyze chain — every resource referenced must have been discovered first.
5. ALWAYS handle empty results gracefully — an empty response is valid data, not an error to retry.

Counter-Rationalizations

Shortcut	Counter	Why
"I'll skip discovery and check known resources"	Always run Phase 1 discovery first	Resource names change, new resources appear — assumed names cause errors
"The user only asked for a quick check"	Follow the full discovery → analysis flow	Quick checks miss critical issues; structured analysis catches silent failures
"Default configuration is probably fine"	Audit configuration explicitly	Defaults often leave logging, security, and optimization features disabled
"Metrics aren't needed for this"	Always check relevant metrics when available	API/CLI responses show current state; metrics reveal trends and intermittent issues
"I don't have access to that"	Try the command and report the actual error	Assumed permission failures prevent useful investigation; actual errors are informative

Common Pitfalls

• Filter specificity: Overly broad filters can match unintended resources -- always dry-run first
• Action ordering: Actions execute in order -- ensure dependent actions are sequenced correctly
• Lambda deployment: Deployed Lambda policies run on schedule -- ensure they have correct IAM permissions
• Cross-account access: Multi-account policies require assumed roles in target accounts
• Output storage: Output directories grow over time -- configure S3 output for long-term storage
• Rate limiting: Policies querying many resources can hit API rate limits -- use max-resources safety
• Mode configuration: Pull mode runs once; Lambda/periodic modes run continuously -- understand the difference