Back to skill

Security audit

YAML

Security checks across malware telemetry and agentic risk

Overview

This YAML helper is coherent and locally scoped, but it keeps persistent local notes about YAML projects and preferences that users should understand before installing.

Install this only if you are comfortable with the skill maintaining local Clawic notes about YAML files, parser/tooling facts, repo style, workflow errors, and preferences across sessions. It claims not to transmit data or store secrets, but local project metadata can still be sensitive in some environments.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Context-Inappropriate Capability

Medium
Confidence
95% confidence
Finding
The documentation instructs the agent to write to persistent skill memory and project artifact files under ~/Clawic/data, which is outside the core YAML authoring/validation scope. In an agent setting, this broadens file-system side effects and can cause unintended persistence, cross-project contamination, or unauthorized modification of unrelated records if followed automatically.

Missing User Warnings

Low
Confidence
86% confidence
Finding
The file directs users to consult and maintain local memory files about pipeline configurations and prior failures, but it does not clearly warn that workflow metadata can contain sensitive operational details such as internal repo structure, validator choices, CI layout, or security-relevant error context. While it explicitly warns not to copy secrets later, the instruction at this location still encourages persistent recording of potentially sensitive project information without data-minimization guidance.

Missing User Warnings

Low
Confidence
93% confidence
Finding
This instruction explicitly tells users to write diagnosed pipeline traps, workflow-file details, and reusable layouts into persistent local files. Even though it says not to copy secret values or masked variables, it still encourages storing verbatim errors and workflow metadata that may expose sensitive project structure, CI architecture, internal tooling, paths, job names, or other operational details useful for reconnaissance.

Missing User Warnings

Medium
Confidence
83% confidence
Finding
The skill instructs the agent to read repo-local and user-home files before making quoting changes, but it does not clearly warn the user that the skill may inspect external files to determine style. In a skill system, hidden file reads and subsequent edits can surprise users, expand the scope of data accessed, and enable unintended modification workflows without explicit consent.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill explicitly directs writing persistent settings to ~/Clawic/data/yaml/config.yaml and memory.md without a clear user-consent warning. Persisting preferences outside the current task can create cross-session state, silently alter future agent behavior, and write to user-home files in a way the user may not expect.

Session Persistence

Medium
Category
Rogue Agent
Content
- ~/clawic/yaml/
---

**Data.** At the start of every session, read `~/Clawic/data/yaml/config.yaml` (what the user declared) and `~/Clawic/data/yaml/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. If none of it exists, work from defaults and say nothing about it.

**Write before the session ends** whenever something durable came out: which parser and spec version a project actually uses; a YAML file that matters and what reads it; a coercion or indentation trap that bit and the fix; the observed house style of an existing repo; a schema, lint config, or file layout that finally validated; a cadence the user agreed to. `memory-template.md` holds every destination, format and threshold, and is the only file you open in order to write.
Confidence
87% confidence
Finding
The skill directs the agent to automatically read and persist session-derived information in local memory files across sessions, including project facts, file inventories, and observed repository style. Even though it says storage is local and excludes credentials, this creates cross-session persistence without explicit per-session consent and can accumulate sensitive operational metadata the user did not intend to retain.

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:54

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
security.md:67