cloud-security-group-audit
Skill1 starsChanged 32 days agoApache-2.0View on GitHub
Audit the network access rules of a public cloud in read-only mode, finding rules open to the whole internet, all-protocol or very wide rules, unattached or unused groups, default groups that carry rules, redundant or shadowed rules, and missing rule descriptions. Use when asked to review, assess, or clean up cloud security groups, network security groups, or VPC firewall rules, or to check what is exposed to the internet. Covers AWS EC2 security groups, Azure network security groups, and Google Cloud VPC firewall rules. Assumes an authenticated, read-capable client for the target cloud account is already available; this skill does not authenticate. It never changes a rule or deletes a group. Checks per platform are in references/INDEX.md.
In this repository40 other files it gives its agents.
- AGENTS.mdAGENTS.md
- alert-monitor-managementSkill
- automation-job-executionSkill
- casb-policy-managementSkill
- casb-posture-assessmentSkill
- certificate-managementSkill
- change-managementSkill
- cmdb-asset-managementSkill
- compute-instance-provisioningSkill
- config-backup-managementSkill
- connector-tunnel-managementSkill
- container-workload-deploymentSkill
- crm-contact-managementSkill
- dcim-asset-managementSkill
- dhcp-scope-managementSkill
- dns-record-managementSkill
- dns-zone-managementSkill
- firewall-managementSkill
- incident-managementSkill
- infrastructure-plan-executionSkill
- interface-link-troubleshootingSkill
- ipam-allocationSkill
- load-balancingSkill
- log-querySkill
- metric-querySkill
- network-access-control-managementSkill
- network-health-verificationSkill
- network-state-collectionSkill
- notificationsSkill
- repository-pull-request-managementSkill
- routing-troubleshootingSkill
- sdwan-policy-managementSkill
- sdwan-site-onboardingSkill
- secret-retrievalSkill
- software-upgrade-lifecycleSkill
- vpn-tunnel-managementSkill
- vulnerability-lookupSkill
- waf-managementSkill
- wireless-network-managementSkill
- zero-trust-access-policy-managementSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 23
Not for creating or editing rules, and not for tracing why one packet was dropped.
- Distinctive: line 53
Read-only. No rule is changed, no group is deleted.
- Very specific: line 54
Only platforms in references/INDEX.md are covered.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Cloud Security Group Audit
Reads the network access rules and group associations of a public cloud and reports where they are exposed, broader than they need to be, unattached, or redundant, each finding backed by the rule or association that proves it.
This skill is diagnostic-shaped: the agent reads state, evaluates it against documented behavior of the platform, and reports findings with evidence. Reading changes nothing, so the procedure has no confirmation gates. The agent does not revoke rules, delete groups, or detach anything; a change of that kind is outside this skill and would need its own gate.
When to use this skill
- Reviewing which rules allow traffic from any address, which groups nothing uses, or which rules overlap, across one account, region, subscription, project, or network.
- Checking whether a specific group or rule exposes a service to the internet.
- Not for creating or editing rules, and not for tracing why one packet was dropped.
Universal concepts
- A rule has a direction, a protocol, a port range, a source or destination, and (where the platform has deny rules) an action. The field names differ by platform.
- A rule is open to the internet when its source covers every address. How "every address" is written differs by platform and is listed in each platform file.
- Some platforms record which resources use a group, while others apply a rule by selector and record no attachment. A check that depends on attachment is defined per platform.
- Where a platform has deny rules it evaluates by priority; where it has only allow rules they combine. Judge exposure and overlap with the platform's own rule order.
- Findings describe state at read time. Report the time of the read and any scope that could not be read (a region, subscription, or project that failed or was not permitted).
Procedure
- Identify the target platform from the request; check
references/INDEX.md. If no row matches, stop: this skill does not cover it. - Identify the scope and the checks requested. Default to all checks across everything the credentials can read.
- Open
references/<platform>/procedure.mdand follow it exactly. Bind each named operation to a concrete call through that platform'sendpoints.md; if no binding is available, report that instead of improvising a call. - Report with the platform's own vocabulary: findings grouped by check, each with the group or rule identifier, the evidence, and the platform's documented reason it matters; then counts per check, then the scopes read and the scopes not read.
Known limitations
- Read-only. No rule is changed, no group is deleted.
- Only platforms in
references/INDEX.mdare covered. - Severity ordering is not defined by the platforms; findings are ordered by check class as the platform file lists them, and what is acceptable for a given workload is the operator's call.
Contributing to this skill
SKILL.md changes only to fix a shared concept above or to add a pointer to a newly supported
platform. Platform detail belongs under references/<platform>/.
Tools
Each platform file names the read operations it uses (list groups, list rules, list associations, and any insights service); the client is the caller's.
Verification checklist
- Platform confirmed against
references/INDEX.md - Every scope (region, subscription, project) read fully, or listed as not read with the reason
- Paged results read to the end
- Each finding carries the rule or association that proves it
- Exposure judged with the platform's own rule order, including deny rules where they exist
- No write call was made
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

