Back to skill

Security audit

Hermes Feishu Guide

Security checks for vulnerabilities and agentic risk

Overview

This is a useful deployment guide, but it tells users to run an unverified remote installer and to configure broad bot access with live secrets.

Review the installer before running it, preferably use a pinned release or checksum-verified script, protect ~/.hermes/.env with restrictive permissions, and avoid open/all-user bot access unless the bot is meant to be available to everyone in the workspace.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

T03 · Remote Payload Retrieval and Execution

Error
Location
SKILL.md:31
Finding

Unverified Remote Installation Script Executed Directly by Shell

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 31
Vulnerability Type: Remote mutable payload retrieval and execution
Risk Level: High

Complete Code Snippet:

bash
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Technical Analysis

The installation command downloads a shell script from the mutable main branch of an external GitHub repository and pipes the response directly into Bash. The downloaded content is not pinned to an immutable commit or release and is not verified using a cryptographic signature or documented checksum.

Consequently, the code executed by users can differ from the code present when this Skill was audited. Direct piping also removes the opportunity to inspect the complete downloaded file before execution. Although the URL points to the apparent upstream Hermes Agent repository, repository ownership alone does not provide payload immutability or protect against account compromise, malicious upstream changes, repository compromise, or an altered network response.

Installing Hermes Agent may legitimately require obtaining upstream software, but immediate execution of mutable remote content exceeds the minimum mechanism necessary. A pinned, verified artifact can provide the same functionality with substantially lower supply-chain risk.

Attack Path

  1. An attacker compromises the upstream repository, a maintainer account, or another component capable of changing main/scripts/install.sh.
  2. The attacker inserts commands into the remote installation script.
  3. A user follows the Skill documentation and runs the provided command.
  4. curl retrieves the attacker-controlled version and streams it directly to Bash.
  5. Bash executes the payload with the permissions of the invoking user.
  6. The payload can read or modify user-accessible files, establish user-level persistence, or collect credentials. If the user in ...[truncated 764 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace the mutable main URL with an artifact pinned to a reviewed release or full commit hash.
  2. Download the installer to a local file rather than streaming it directly into a shell:
    bash
    curl -fL -o install.sh \
      https://raw.githubusercontent.com/NousResearch/hermes-agent/<FULL_COMMIT_HASH>/scripts/install.sh
    
  3. Publish and verify an authenticated signature where supported. At minimum, provide a trusted SHA-256 digest and verify it before execution:
    bash
    echo '<EXPECTED_SHA256>  install.sh' | sha256sum --check -
    
  4. Instruct the user to inspect the downloaded script before running it.
  5. Execute it explicitly and without elevated privileges:
    bash
    bash install.sh
    
  6. Document the files, network destinations, and configuration changes the installer is expected to create so unexpected behavior can be identified.
  7. Require restrictive permissions on the secrets file, for example:
    bash
    chmod 600 ~/.hermes/.env
    
  8. Prefer FEISHU_GROUP_POLICY=allowlist and disable GATEWAY_ALLOW_ALL_USERS unless unrestricted access is an explicit operational requirement.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (5)

Chaining Abuse

High
Category
Tool Misuse
Confidence
97% confidence
Finding

The | bash construct is a classic chaining-abuse pattern because it turns unreviewed network data into executable shell commands in a single step. This is especially risky in a local deployment guide, where users may run the command with their own account context and immediately expose the machine to arbitrary code execution.

Content

Scanner excerpt · SKILL.md (reported line 31)May include surrounding context.

一键安装

bash
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
source ~/.zshrc

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The guide instructs users to execute a remotely fetched script directly with bash, which removes the opportunity to inspect, pin, or verify the script before execution. If the upstream repository, network path, or referenced branch is compromised, this becomes arbitrary code execution on the user's local machine.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill tells users to place live API keys and app secrets in a local .env file but does not warn about file permissions, accidental commits, log leakage, or shell/history exposure. This increases the chance that credentials are stored insecurely and later exfiltrated, enabling unauthorized access to the bot platform and LLM provider.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The recommended configuration combines FEISHU_GROUP_POLICY=open with GATEWAY_ALLOW_ALL_USERS=true, effectively allowing broad, potentially unrestricted interaction with the bot. In a messaging bot context this can expose internal capabilities or sensitive data to unauthorized users, increase prompt-injection exposure, and enable abuse of connected tools or APIs.

Content

No source excerpt is available for this finding.

External Script Fetching

Low
Category
Supply Chain
Confidence
94% confidence
Finding

Fetching and immediately executing an external script from a remote URL is dangerous because the content can change without notice and is trusted implicitly at runtime. In this deployment guide context, users are likely to run the command verbatim, so compromise of the source or delivery path would directly compromise the host.

Content

Scanner excerpt · SKILL.md (reported line 31)May include surrounding context.

一键安装

bash
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
source ~/.zshrc