Back to skill

Security audit

Awn

Security checks for vulnerabilities and agentic risk

Overview

The skill’s messaging purpose is coherent, but its install instructions tell users to execute a mutable remote shell script directly, which warrants Review before installation.

Review this skill carefully before installing. Prefer downloading a pinned release artifact and verifying its checksum or signature instead of running the documented curl-to-bash installer. If you use AWN, understand that it creates a local agent identity, stores data under ~/.awn, starts a background daemon, contacts a Gateway, and listens for peer connections when running.

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:21
Finding

Mutable Remote Installer Is Downloaded and Executed Without Integrity Verification

Content
View full analysis

Vulnerability Details

File Locations:

  • SKILL.md:21
  • references/install.md:6
  • references/install.md:12
  • references/install.md:18
  • references/flows.md:6

Vulnerability Type: Remote payload retrieval and direct shell execution
Risk Level: High

Vulnerable Code — SKILL.md:21:

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Vulnerable Code — references/install.md:6:

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Vulnerable Code — references/install.md:12:

bash
INSTALL_DIR=/usr/local/bin curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Vulnerable Code — references/install.md:18:

bash
VERSION=x.y.z curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Vulnerable Code — references/flows.md:6:

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Technical Analysis

The installation instructions retrieve install.sh from the mutable main branch of an external GitHub repository and stream its contents directly into Bash. The downloaded script is not included in the audited project, pinned to an immutable commit, or checked against a trusted checksum or cryptographic signature.

Consequently, the code executed by users can change after this Skill has been reviewed. HTTPS protects the connection in transit, but it does not protect against compromise of the upstream repository, maintainer account, release process, or intentional later modification of the script. Piping directly to Bash also prevents meaningful inspection before execution.

The remote script runs ...[truncated 2429 chars]

Remediation
View remediation

Remediation Suggestions

  1. Remove every curl ... | bash installation instruction.
  2. Publish versioned CLI artifacts through immutable releases and reference a specific release version rather than the mutable main branch.
  3. Publish SHA-256 checksums through a separately protected release channel, or cryptographically sign release artifacts using a verifiable signing identity.
  4. Require users to download the archive first, verify its checksum or signature, and only then extract and install it:
    bash
    curl -fL -o awn.tar.gz "https://github.com/ReScienceLab/agent-world-network/releases/download/v1.6.0/awn-v1.6.0-x86_64-unknown-linux-gnu.tar.gz"
    echo "<trusted-sha256>  awn.tar.gz" | sha256sum --check -
    tar -xzf awn.tar.gz
    install -m 0755 awn-v1.6.0/awn "$HOME/.local/bin/awn"
    
  5. If an installer script remains necessary, pin its URL to an immutable commit hash, download it as a file, verify a published checksum or signature, and instruct the user to inspect it before execution.
  6. Default installation to a user-owned directory such as ~/.local/bin. Document system-wide installation separately and explain when elevated privileges are required.
  7. Keep the Skill version synchronized with a fixed CLI release so that the audited documentation cannot silently install a materially different future version.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (14)

Chaining Abuse

High
Category
Tool Misuse
Confidence
98% confidence
Finding

The | bash construct is the dangerous sink: it turns untrusted network input into shell commands with no opportunity for review, sandboxing, or integrity verification. In a skill intended for AI-agent operators, this is more dangerous because users may copy/paste automation-oriented instructions into privileged environments, amplifying the effect of a compromised installer.

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

Install

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Installs the latest release to ~/.local/bin/awn. Set INSTALL_DIR to override.

Chaining Abuse

High
Category
Tool Misuse
Confidence
99% confidence
Finding

The | bash pattern is a direct command-chaining mechanism that turns remote content retrieval into immediate shell execution, removing any opportunity for validation before code runs. In this context, the danger is amplified because the project is an agent networking CLI that may later connect to external peers, making initial compromise especially costly.

Content

Scanner excerpt · references/flows.md (reported line 6)May include surrounding context.

Flow 1 — First-time setup

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash
awn daemon start
awn status

Chaining Abuse

High
Category
Tool Misuse
Confidence
98% confidence
Finding

The | bash pipeline is the core unsafe behavior: it chains network retrieval directly into command execution with no inspection boundary. In documentation for an install flow, this normalizes a risky pattern and creates a single-step remote code execution path if the fetched content is altered.

Content

Scanner excerpt · references/install.md (reported line 6)May include surrounding context.

One-liner

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Installs to ~/.local/bin/awn. Set INSTALL_DIR to change the destination:

Chaining Abuse

High
Category
Tool Misuse
Confidence
98% confidence
Finding

This command combines direct execution of network content with a likely system-wide install target, making exploitation more damaging. If the fetched script is malicious, the chaining pattern can lead to unauthorized filesystem changes or broader system compromise depending on how the install is run.

Content

Scanner excerpt · references/install.md (reported line 12)May include surrounding context.

Installs to ~/.local/bin/awn. Set INSTALL_DIR to change the destination:

bash
INSTALL_DIR=/usr/local/bin curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Specific version

Chaining Abuse

High
Category
Tool Misuse
Confidence
97% confidence
Finding

Even with a version environment variable, piping remote content into bash remains an unsafe command-chaining pattern because trust is placed in live network content at execution time. The danger is amplified by the product context: this is an agent-networking CLI, so users are being asked to install software that will later participate in networked messaging and daemonized behavior.

Content

Scanner excerpt · references/install.md (reported line 18)May include surrounding context.

Specific version

bash
VERSION=x.y.z curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

From GitHub Release

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The installation instruction pipes a remotely fetched script directly into bash, which gives whatever content is served at that URL immediate code execution on the user's system. Even if the upstream project is legitimate, compromise of the GitHub account, repository, branch, CDN path, or network trust chain could turn this into a supply-chain attack with no inspection step by the user.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documentation instructs users to fetch and execute a remote installer and then immediately start a network-connected daemon without any warning about trust, code review, or the risks of connecting to external worlds and agents. In an agent/networking tool, this lowers user caution and can lead to execution of changed upstream code or unintended exposure to untrusted peers.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The documentation instructs users to fetch and immediately execute a remote shell script from a mutable GitHub branch via curl | bash without any integrity verification, review step, or risk warning. If the upstream repository, branch, network path, or hosting account is compromised, users could execute arbitrary code on their systems during installation.

Content

No source excerpt is available for this finding.

External Script Fetching

Low
Category
Supply Chain
Confidence
96% confidence
Finding

This skill explicitly tells users to fetch and execute an external script from GitHub, creating a direct trust dependency on mutable remote content. Because the command targets main rather than a fixed version and immediately executes the result, any malicious or accidental upstream change would execute locally without further validation.

Content

Scanner excerpt · SKILL.md (reported line 21)May include surrounding context.

Install

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Installs the latest release to ~/.local/bin/awn. Set INSTALL_DIR to override.

External Script Fetching

Low
Category
Supply Chain
Confidence
97% confidence
Finding

The example uses curl ... | bash, which executes a remotely hosted script directly from GitHub without review, pinning, or integrity verification. If the source is compromised, the URL content changes, or a user is subject to network or supply-chain compromise, arbitrary code runs immediately on the host.

Content

Scanner excerpt · references/flows.md (reported line 6)May include surrounding context.

Flow 1 — First-time setup

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash
awn daemon start
awn status

External Script Fetching

Low
Category
Supply Chain
Confidence
96% confidence
Finding

This command fetches an external script from GitHub and executes it immediately in a shell. Because the content is remote and not integrity-checked, any compromise of the source or delivery path can result in arbitrary command execution on the user's machine.

Content

Scanner excerpt · references/install.md (reported line 6)May include surrounding context.

One-liner

bash
curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Installs to ~/.local/bin/awn. Set INSTALL_DIR to change the destination:

External Script Fetching

Low
Category
Supply Chain
Confidence
96% confidence
Finding

This variant still executes a remotely fetched script without validation, while additionally encouraging installation into /usr/local/bin, which commonly requires elevated privileges. That combination can increase the consequences of malicious script execution by affecting system-wide binaries or prompting unsafe privilege escalation habits.

Content

Scanner excerpt · references/install.md (reported line 12)May include surrounding context.

Installs to ~/.local/bin/awn. Set INSTALL_DIR to change the destination:

bash
INSTALL_DIR=/usr/local/bin curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

Specific version

External Script Fetching

Low
Category
Supply Chain
Confidence
96% confidence
Finding

Although this example pins a version variable, it still retrieves and executes a remote script directly rather than verifying an immutable artifact. Version selection alone does not protect against tampering with the install script or the referenced source.

Content

Scanner excerpt · references/install.md (reported line 18)May include surrounding context.

Specific version

bash
VERSION=x.y.z curl -fsSL https://raw.githubusercontent.com/ReScienceLab/agent-world-network/main/packages/awn-cli/install.sh | bash

From GitHub Release

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The post-install instructions tell the user to run awn daemon start and only comment that it starts a background daemon, but do not explain what privileges, persistence, network behavior, or resource usage the background process entails. Starting an always-on networked daemon without clear disclosure can surprise users and increases operational and privacy risk in an agent-to-agent messaging tool.

Content

No source excerpt is available for this finding.