Back to skill

Security audit

Deepvista Shared

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a DeepVista CLI reference, but it tells agents to silently run an update-check command whenever the skill loads.

Review this before installing if you do not want skills to run commands merely by being loaded. The DeepVista CLI integration itself is coherent, but the silent update check should be removed or made explicitly user-approved, and users should prefer isolated, non-privileged installation with a trusted, pinned package version where possible.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (2)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:18
Finding

Silent Automatic Command Execution When the Skill Is Loaded

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
SKILL.md:9
Finding

Unpinned Third-Party CLI Installation Creates Supply-Chain Risk

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 184)May include surrounding context.

md
2. **Read-only commands** are safe to run without confirmation.
3. **Never output tokens or secrets** — use `deepvista auth status` to check auth state.
4. **Use `--dry-run`** to preview destructive operations before executing.
5. **Tokens are sensitive** — stored in `~/.config/deepvista/credentials.json` (mode 0600).

## See Also

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs the agent to automatically execute deepvista skill check-update on load, even though the skill is presented as documentation for authentication and conventions. This creates hidden command execution and likely network activity without explicit user consent, which violates least surprise and expands the skill's behavior beyond passive reference material.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

A silent background update check is not necessary for the declared purpose of a shared authentication/security reference skill. Hidden execution of a network-capable CLI increases attack surface, can leak environmental metadata, and conditions agents to perform side effects unrelated to the user's request.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The background command is meant to run silently and suppress failures, despite invoking a CLI that may perform network communication. Silent execution prevents meaningful user awareness or consent and makes it harder to audit what actions the agent took on the system.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · SKILL.md (reported line 180)May include surrounding context.

md
## Security Rules

1. **Write commands** are marked with `> [!CAUTION]` — always confirm with the user before executing write/delete operations.
2. **Read-only commands** are safe to run without confirmation.
3. **Never output tokens or secrets** — use `deepvista auth status` to check auth state.
4. **Use `--dry-run`** to preview destructive operations before executing.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 181)May include surrounding context.

md
## Security Rules

1. **Write commands** are marked with `> [!CAUTION]` — always confirm with the user before executing write/delete operations.
2. **Read-only commands** are safe to run without confirmation.
3. **Never output tokens or secrets** — use `deepvista auth status` to check auth state.
4. **Use `--dry-run`** to preview destructive operations before executing.
5. **Tokens are sensitive** — stored in `~/.config/deepvista/credentials.json` (mode 0600).