Back to skill

Security audit

Blackbox

Security checks for vulnerabilities and agentic risk

Overview

This skill coherently delegates coding work to Blackbox AI, but it asks users to install and trust a mutable third-party CLI with broad coding-agent authority without enough safety guidance.

Review this carefully before installing. Use a pinned, reviewed Blackbox CLI version in an isolated environment, avoid `--yolo` on real repositories, do not run install commands with elevated privileges, and assume prompts and code may be sent to Blackbox or configured model providers. Remove secrets from projects before delegation and review all file changes before merging or running 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:24
Finding
Unpinned Third-Party CLI Installation Creates a Supply-Chain Risk<![CDATA[ ## Vulnerability Details **File Location**: `SKILL.md`, lines 24-29 **Vulnerability Type**: Unpinned and globally installed third-party dependencies **Risk Level**: Medium ### Vulnerable Code Snippet ```markdown - Blackbox CLI installed: `npm install -g @blackboxai/cli` - Or install from source: ``` git clone https://github.com/blackboxaicode/cli.git cd cli && npm install && npm install -g . ``` ``` ### Technical Analysis The installation instructions retrieve and execute third-party content without pinning it to a reviewed package version, integrity hash, release artifact, or source commit. The command `npm install -g @blackboxai/cli` resolves the package version at installation time and may execute package lifecycle scripts with the permissions of the invoking user. Its behavior can therefore change after this Skill has been reviewed. Global installation also makes the resulting executable available broadly within the user's environment. The alternative source installation clones the current default branch of an external repository and installs both that mutable source and its dependency graph. Because no commit or signed release is specified, the installed code may differ between executions. This finding concerns the documented installation process. The audited project itself contains only `SKILL.md`; no bundled executable or malicious script was found. ### Attack Path 1. An attacker compromises the referenced npm package, its publishing account, the source repository, or a transitive dependency. 2. The attacker publishes malicious package content, modifies the repository's default branch, or introduces a malicious lifecycle script. 3. A user follows the documented prerequisite and runs one of the unpinned installation procedures. 4. npm downloads the attacker-controlled content and may execute its lifecycle scripts during installation. 5. The malicious code executes with the permissions of the user running npm and can install or ...[truncated 877 chars]
Remediation
<![CDATA[ ## Remediation Suggestions 1. Pin the npm dependency to a specific reviewed version rather than resolving the latest available release: ```sh npm install -g @blackboxai/cli@<reviewed-version> ``` 2. Document package integrity verification using a trusted checksum, registry integrity metadata, or signed release artifact. 3. For source installation, pin the repository to a reviewed commit or signed tag: ```sh git clone https://github.com/blackboxaicode/cli.git cd cli git checkout <reviewed-commit-hash> ``` 4. Use lockfile-enforced dependency installation, such as `npm ci`, with a reviewed and committed lockfile. 5. Review package lifecycle scripts before installation and disable them with `--ignore-scripts` when they are not required. 6. Prefer an isolated, non-global installation environment, such as a dedicated container or restricted user account, to limit access to host files and credentials. 7. Explicitly warn users not to run the installation as root or through `sudo`. 8. Establish a process for periodically reviewing pinned versions and updating them only after source, dependency, and release-integrity validation. ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Memory PoisoningPersistent Context Injection, Context Window Stuffing, Memory Manipulation
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Memory Manipulation

High
Category
Memory Poisoning
Content
| Command | Effect |
|---------|--------|
| `/compress` | Shrink conversation history to save tokens |
| `/clear` | Wipe history and start fresh |
| `/stats` | View current token usage |
| `Ctrl+C` | Cancel current operation |
Confidence
80% confidence
Finding
Skill manipulates agent memory, state, or stored context. Memory corruption can alter personality, override safety rules, or cause unpredictable behavior.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill instructs users to configure a Blackbox API key for a third-party CLI but gives no guidance on secure storage, scope minimization, logging exposure, or the privacy implications of sending code and prompts to an external service. Because this skill is specifically designed to delegate coding tasks, users may inadvertently expose proprietary source code or credentials to the provider.

Autonomous Decision Making

Medium
Category
Excessive Agency
Content
|------|--------|
| `--prompt "task"` | Non-interactive one-shot execution |
| `--resume-checkpoint "tag"` | Resume from a saved checkpoint |
| `--yolo` | Auto-approve all actions and model switches |
| `blackbox session` | Start interactive chat session |
| `blackbox configure` | Change settings, providers, models |
| `blackbox info` | Display system information |
Confidence
93% confidence
Finding
`--yolo` explicitly enables autonomous decision-making by allowing the delegated agent to approve actions and model switches without human oversight. For a multi-model coding agent operating through a terminal, this can translate into unreviewed modifications, command execution, or risky environment interactions.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill advertises `--yolo` as an available flag that auto-approves all actions and model switches, but does not warn that this can authorize filesystem changes, code execution, or other destructive operations without user review. In the context of a coding agent that delegates work to a third-party CLI, normalizing auto-approval materially increases the chance of unsafe autonomous behavior.

Static analysis

No suspicious patterns detected.