Back to skill

Security audit

SelfEnvolveEngine

Security checks for vulnerabilities and agentic risk

Overview

This skill is a broad Prismer Cloud integration guide that is not clearly malicious, but it asks agents to install unpinned packages and use cloud messaging, memory, file, and skill-install features with limited safety boundaries.

Install only if you trust Prismer Cloud and are comfortable sending selected content to its services. Pin and review package versions where possible, avoid running unpinned npx in privileged environments, do not upload secrets or regulated data without explicit approval, and manually inspect any skills installed from Prismer before letting an agent use them.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • 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 (1)

T08 · Insecure Dependencies

Warning
Location
Skill.md:13
Finding

Unpinned Third-Party npm Packages Are Installed and Executed

Content
View full analysis

Vulnerability Details

File Location: Skill.md, lines 13, 277, and 329
Vulnerability Type: Unpinned third-party dependency installation and execution
Risk Level: Medium

Vulnerable Code

bash
# Skill.md:13
which prismer || npm install -g @prismer/sdk
markdown
<!-- Skill.md:277 -->
| **MCP Server** | `npx -y @prismer/mcp-server` (33 tools) |
markdown
<!-- Skill.md:329 -->
| MCP Server | `@prismer/mcp-server` | `npx -y @prismer/mcp-server` (33 tools) |

Technical Analysis

The documented commands resolve npm packages without pinning exact, previously reviewed versions or requiring integrity and provenance verification.

npm install -g @prismer/sdk installs the currently resolved package globally and may execute npm lifecycle scripts with the invoking user's privileges. The global installation also modifies the user's persistent tooling environment.

npx -y @prismer/mcp-server is more direct: it automatically accepts installation, downloads the currently resolved package release, and executes its entry point. Because no version is pinned, the code executed by users may differ from the code that was previously audited. The documented MCP server exposes 33 tools, potentially increasing the resources available to a compromised process.

This is a supply-chain weakness rather than evidence that the named packages are currently malicious. Exploitation would require package-publisher compromise, registry compromise, publication of a malicious release, or another dependency-resolution attack.

Attack Path

  1. An attacker compromises the npm publisher account, package distribution process, registry path, or a transitive dependency.
  2. The attacker publishes a malicious release under @prismer/sdk or @prismer/mcp-server.
  3. A user or agent follows Skill.md and runs the unpinned installation or npx -y command.
  4. npm resolves the attacker-controlled release becau ...[truncated 1081 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin every package to an exact reviewed version, for example @prismer/sdk@X.Y.Z and @prismer/mcp-server@X.Y.Z; do not use ranges or implicit latest resolution.
  2. Replace npx -y with a locally declared, lockfile-pinned dependency. Execute the local binary only after dependency review and installation.
  3. Avoid global installation. Prefer a project-local installation in a dedicated directory or isolated environment.
  4. Commit and enforce an npm lockfile with integrity hashes. Use npm ci in automated environments so dependency resolution cannot drift silently.
  5. Verify package provenance, publisher identity, registry source, signatures or attestations, and expected integrity before execution.
  6. Review direct and transitive dependencies and monitor them for publisher changes, unexpected releases, known vulnerabilities, and suspicious lifecycle scripts.
  7. Require explicit user approval before downloading or executing packages. Do not automatically approve package installation in privileged agent workflows.
  8. Run the MCP server with least privilege in a sandbox or container. Restrict filesystem access, environment variables, credentials, network egress, and exposed tools to those strictly required.
  9. Document trusted package versions and a controlled upgrade procedure that includes source review, integrity verification, testing, and rollback.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Rogue AgentSelf-Modification, Session Persistence
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Self-Modification

High
Category
Rogue Agent
Confidence
94% confidence
Finding

The skill explicitly supports installing additional skills that are written locally as SKILL.md, which is a self-modification and trust-boundary expansion mechanism. In an agent environment, this is dangerous because a remotely sourced skill can alter future agent behavior, introduce prompt-injection content, or add new risky instructions without prior review.

Content

Scanner excerpt · Skill.md (reported line 217)May include surrounding context.

bash
prismer skill find "evolution"                      # search catalog
prismer skill find -c repair                        # filter by category
prismer skill install <slug>                        # install + write SKILL.md locally
prismer skill list                                  # installed skills
prismer skill show <slug>                           # view skill content
prismer skill uninstall <slug>                      # uninstall

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill encourages use of web loading, search, document parsing, webhooks, SSE, and related cloud-backed features without clearly warning that URLs, documents, messages, and tokens may be transmitted to external services. In a security-sensitive agent context, that increases the risk of privacy leaks, accidental exfiltration of proprietary content, and unsafe handling of sensitive documents.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

The skill documents commands that send messages, edit or delete conversations, upload files, and write or delete persistent memory, but it does not prominently warn that these operations mutate remote or persistent state. In an agent setting, this omission can lead to unintended disclosure, destruction, or modification of user data because the skill normalizes state-changing actions without consent boundaries.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The skill recommends executing an MCP server package via npx without pinning a specific version. This creates a supply-chain risk because future package updates, dependency compromise, or namespace hijacking could cause agents to run unexpected code with the agent's local privileges.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This is a second instance of the same unpinned execution pattern, again instructing users to run a remote package at whatever version is current. Repeated exposure increases the chance that an agent will directly execute unreviewed code from the registry.

Content

No source excerpt is available for this finding.