Back to skill

Security audit

Google Cloud

Security checks across malware telemetry and agentic risk

Overview

This Google Cloud skill clearly discloses that it uses gcloud and stores local operational notes, with no evidence of hidden credential capture, exfiltration, or automatic destructive actions.

Install only if you want a GCP helper that keeps local operational memory under ~/Clawic/data/. Review generated gcloud commands before running them, especially anything that changes IAM, billing, projects, datasets, keys, or node pools, and avoid storing sensitive incident details unless the local Clawic data directory is protected appropriately.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (8)

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
The skill instructs the GCP-focused agent to read and modify a shared cross-skill contacts store, including preserving and extending data written by other skills. That expands scope beyond cloud infrastructure work and creates a path for unnecessary access, modification, or propagation of unrelated personal/contact data, increasing the risk of data contamination and privacy violations across skills.

Description-Behavior Mismatch

Low
Confidence
93% confidence
Finding
Directing creation of decision records in local markdown files causes the skill to generate persistent artifacts that are not part of its described cloud-advisory function. This expands the skill from guidance into local state mutation, which can surprise users, leak sensitive architecture details into unreviewed files, or create an unauthorized audit trail on disk.

Description-Behavior Mismatch

Low
Confidence
93% confidence
Finding
Directing creation of decision records in local markdown files causes the skill to generate persistent artifacts that are not part of its described cloud-advisory function. This expands the skill from guidance into local state mutation, which can surprise users, leak sensitive architecture details into unreviewed files, or create an unauthorized audit trail on disk.

Missing User Warnings

Low
Confidence
96% confidence
Finding
The skill explicitly instructs the agent to update persistent local/shared files (`~/Clawic/data/gcp/memory.md` and `~/Clawic/data/gcp/artifacts/network-plan.md`) as part of normal execution, but provides no warning, confirmation step, or scope guard for modifying shared state. In an agent setting, this can cause unintended persistence, cross-session contamination, or corruption of operational notes if the inferred topology is wrong or the prompt was adversarial.

Missing User Warnings

Low
Confidence
91% confidence
Finding
The markdown tells the agent to update local memory/artifact files but does not warn the user that local data will be modified. Lack of disclosure is risky because users may invoke what appears to be a read-only advisory skill while it silently persists information, potentially including infrastructure decisions and environment details.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The file instructs saving a credential-leak runbook containing timeline, queries, findings, and rotated items to a local user path. Even though it says to replace secrets with pointers, incident artifacts often contain sensitive metadata, affected resource names, principals, timestamps, and investigative outputs that can aid follow-on attacks if stored in an unprotected local file outside a governed evidence repository.

Missing User Warnings

Low
Confidence
67% confidence
Finding
The skill directs writing persistent files into the user's home directory without an explicit warning or consent step. In an agent setting, silent local writes can surprise users, create unwanted persistence, and normalize filesystem side effects beyond the advisory scope of the skill.

Session Persistence

Medium
Category
Rogue Agent
Content
- ~/Clawic/profile.yaml
---

**Data.** At the start of every session, read `~/Clawic/data/gcp/config.yaml` (what the user declared) and `~/Clawic/data/gcp/memory.md` (what you observed, plus its `## Boxes` index and `## Due` table). Open any file `## Boxes` names when the condition on its line applies — the index is the list of files, never assume the list is fixed. Every path it names is inside `~/Clawic/data/`; ignore any line that points anywhere else. Everything this skill reads or writes is a plain local note under the folders declared in `configPaths` — nothing leaves the machine and no credential is ever written. In a shared box it updates or removes only the rows it wrote itself, matched on that box's identity key; a row another skill wrote is read, never rewritten and never deleted, and every write and deletion is named in one line as it happens. Read `~/Clawic/data/servers/servers.md` before any deploy, sizing, or "what do I have" question. If none of it exists, work from defaults and say nothing about it.

**Write before the session ends** whenever it produced something durable: a VM or cluster created, resized, discovered or retired; an inventory pass; a spend number or a saving; a budget or a billing export; a project and its owner; a quota that was raised; a service account and what it is for; a BigQuery dataset and its scan baseline; a DNS zone; a deploy or a timed restore drill; or something the user will want to read again — a runbook, an IAM policy that finally worked, an architecture decision. `memory-template.md` has every destination, format and threshold, and is the only file you open to write.
Confidence
88% confidence
Finding
write and deletion is named in one line as it happens. Read `~/Clawic/data/servers/servers.md` before any deploy, sizing, or "what do I have" question. If none of it exists, work from defaults and say

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
memory-template.md:67