CC deploying-cloud-deception-with-decoy-resources
Deploy cloud-native deception across AWS, Azure, and GCP using decoy (honey) resources whose only purpose is to generate a high-fidelity alert the instant an attacker touches them: canary IAM access keys, permission-less decoy users/roles/service principals, honey object-storage buckets, and decoy secrets in Secrets Manager / Key Vault / Secret Manager. Wires detection through CloudTrail + EventBridge, Azure Sentinel honeytoken watchlists + Defender, and GCP Cloud Audit Logs, so any use of a decoy is routed to the SOC with near-zero false positives. Use when protecting cloud accounts and data stores, when an org has only on-prem honeypots and needs cloud coverage, when seeding fake AWS keys to catch credential theft and code-leak exposure, or when detecting cloud reconnaissance and lateral movement. Keywords: cloud deception, canary token AWS, honey S3 bucket, decoy IAM credentials, CloudTrail alert, GuardDuty, Sentinel honeytoken, decoy secret, honey service account, cloud honeypot, breach detection.
Deploy cloud-native deception across AWS, Azure, and GCP using decoy (honey) resources whose only purpose is to generate a high-fidelity alert the instant an…
As a process C 62/100 · Has gaps — weak spots: when it triggers, failures and branches, running it twice
How to improve
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 9
✓ No critical or high findings
Medium and low: 9
-
low Risky intent
intent-offensive-securityassets/template.md:24Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| svc-backup-prod | 2026-05-20 | Red team | 90 s | PASS |
-
low Risky intent
intent-offensive-securityassets/template.md:25Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| acme…026 | 2026-05-20 | Red team | 2 min | PASS |
-
low Risky intent
intent-offensive-securityassets/template.md:26Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| prod/db/master-password | 2026-05-20 | Red team | 75 s | PASS |
-
low Risky intent
intent-offensive-securityassets/template.md:27Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| Azure honeytoken SP | 2026-05-20 | Red team | 6 min | PASS |
-
low Risky intent
intent-offensive-securityassets/template.md:28Offensive-security / dual-use content (legitimate for authorised testing; review intended use)| svc-billing-export | 2026-05-20 | Red team | 4 min | PASS |
-
low Risky intent
intent-offensive-securitySKILL.md:13Offensive-security / dual-use content (legitimate for authorised testing; review intended use)reconnaissance and lateral movement. Keywords: cloud deception, canary token AWS, honey
-
low Risky intent
intent-offensive-securitySKILL.md:51Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)- When detecting cloud reconnaissance (enumeration of IAM, storage, or secrets) and lateral movement that legitimate users would never perform.
detector -
low Risky intent
intent-offensive-securitySKILL.md:67Offensive-security / dual-use content (legitimate for authorised testing; review intended use)Pick decoys that match how *your* attackers operate: leaked AWS keys (credential theft), an "admin" S3 bucket (data discovery), a `prod-db-password` secret (secrets harvesting), a privileged-looking s
-
low Risky intent
intent-offensive-securitySKILL.md:155Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- **Cloud lateral movement.** A permission-less decoy service principal that "looks" privileged catches adversaries assuming roles during pivoting.
Files scanned: 5. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- note
frontmatter-keyunknown frontmatter key "domain" - note
frontmatter-keyunknown frontmatter key "subdomain" - note
frontmatter-keyunknown frontmatter key "nist_csf" - note
frontmatter-keyunknown frontmatter key "mitre_attack"
Process rating: all ten parameters 62/100
- 0Failures and branches. Linear process with no failure handling
- 20When it triggers. No condition that starts the skill
- 30Running it twice. 14 mutating operations with no state check
- 60Tools and files. Uses tools (web) that frontmatter does not declare
- 60Result and completion. Output format stated, no completion criterion
- 70Inputs and preconditions. Inputs and preconditions are listed
- 100Steps. 25 steps
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 2438 tokens
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +4Description does not say when NOT to use the skill (false activations)
- +3Description length 1016: 120–800 characters recommended
- -41 reference files, but SKILL.md never points to them: the model will not open them
- +2Single-language instructions
- +4Structure: 18 headings
- +3Step-by-step instructions: 25 items
- +3Output format is stated explicitly
- +4Has examples (5 code blocks)
- +3All 1 scripts are documented
- +1License stated
Quality base 70; lint remarks subtract, signals add up to 100. Result: 81.