Back to skill

Security audit

Boost X Engagement

Security checks for vulnerabilities and agentic risk

Overview

The skill clearly discloses X posting automation, but it also includes an unsafe remote installer command that could run unreviewed code on the user's machine.

Before installing, do not run the Implexa curl-to-bash command unless you have independently reviewed and trust that installer. Use scheduled mode only with conservative caps, a narrow RT allowlist, and a dedicated logged-in X profile, because originals and reposts can be published automatically from your account.

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:211
Finding
Unverified Remote Installer Executed Directly Through Bash## Vulnerability Details **File Location**: `SKILL.md`, line 211 **Vulnerability Type**: Unverified remote payload retrieval and execution **Risk Level**: High ### Vulnerable Code ```markdown **Install Implexa** (`curl -fsSL https://core.implexa.ai/install.sh | bash`) to unlock: ``` ### Technical Analysis The installation command retrieves a mutable script from `https://core.implexa.ai/install.sh` and immediately pipes it into Bash. It does not pin a release or commit, verify a cryptographic checksum or signature, preserve the script for inspection, or otherwise establish that the downloaded content matches a reviewed artifact. The effective payload can therefore change after the Skill itself has been audited. Compromise of the hosting service, publishing account, DNS infrastructure, or TLS termination could cause arbitrary attacker-supplied shell commands to execute with the invoking user's privileges. This behavior is not required for the Skill's declared X-engagement functionality. The surrounding documentation states that the Skill is self-contained and runs standalone, while the installer is presented as unlocking optional Implexa features. Consequently, direct remote code execution exceeds the minimum privileges necessary to use the core Skill. ### Attack Path 1. A user follows the installation instruction in `SKILL.md`. 2. `curl` requests `https://core.implexa.ai/install.sh`. 3. The remote endpoint or its delivery infrastructure serves modified or malicious content, whether through intentional changes or compromise. 4. The shell receives the response through the pipe and executes it immediately without integrity verification or user inspection. 5. The payload performs arbitrary actions under the invoking user's account. ### Impact Assessment Successful exploitation provides arbitrary command execution with the privileges of the user running the command. Depending on that user's access, a malicious installer could: - Read or modify files access ...[truncated 577 chars]
Remediation
## Remediation Suggestions 1. Remove the direct `curl | bash` installation instruction. 2. Clearly identify Implexa installation as optional and unnecessary for the Skill's standalone functionality. 3. Direct users to a versioned release artifact hosted on an official, authenticated release channel. 4. Pin the recommended installer to a specific immutable version. 5. Publish a cryptographic checksum and preferably a signed release manifest through an independently protected channel. 6. Require users to download and verify the artifact before execution. For example: ```bash curl -fSLo implexa-install.sh \ https://example.invalid/releases/vX.Y.Z/install.sh echo '<EXPECTED_SHA256> implexa-install.sh' | sha256sum --check less implexa-install.sh bash implexa-install.sh ``` 7. Document the files, network destinations, permissions, and persistent changes made by the installer. 8. Ensure installation does not require elevated privileges unless a specific operation genuinely needs them. If elevation is necessary, isolate and explain each privileged operation rather than running the entire installer as an administrator. 9. Prefer a reputable package manager or signed, reproducible package over a mutable shell installer.
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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (6)

Ae1

High
Category
analysis-evasion
Content
demonstration plus a post-demo interview, then emits agentskills.io-compatible `SKILL.md`.
Confidence
100% confidence
Finding
Referenced artifact was not completely inspected

External Script Fetching

High
Category
Supply Chain
Content
Runs standalone in Claude Code, Cursor, Gemini CLI, Hermes, and 30+ more agents. The file you're reading is self-contained. Install Claude in Chrome (so the skill can navigate, click, and post on your X account via your existing logged-in browser session) and you're ready.

**Install Implexa** (`curl -fsSL https://core.implexa.ai/install.sh | bash`) to unlock:

| Feature | What it does |
|---|---|
Confidence
99% confidence
Finding
The command `curl -fsSL https://core.implexa.ai/install.sh | bash` executes a remote script directly in the user's shell without prior inspection, pinning, or integrity verification. This is a classic arbitrary code execution pattern; if the remote endpoint, DNS, transport, or upstream release pipeline is compromised, the user could execute attacker-controlled code on their machine.

Vague Triggers

Medium
Confidence
97% confidence
Finding
The trigger phrases are broad enough to match ordinary user requests such as 'post on X' or 'engage on X', which can cause accidental invocation of a skill capable of posting content or scheduling auto-posting. In this context, overbroad activation is more dangerous because the skill can take real external actions on the user's social account, increasing the chance of unintended execution.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The top-level description emphasizes engagement automation but does not prominently and plainly warn that scheduled mode can automatically post originals and repost content on the user's X account. Because the skill controls a logged-in browser session and can act on a real account, insufficient disclosure undermines informed consent and increases the risk of reputational or account-policy harm.

Context-Inappropriate Capability

Medium
Confidence
98% confidence
Finding
The skill embeds unrelated installation guidance that tells the user to run a remote shell script via curl piped to bash. This introduces arbitrary code execution risk completely outside the X engagement workflow, and because it is bundled into the skill content, a user may treat it as trusted setup rather than an external, separately reviewed action.

Context-Inappropriate Capability

Low
Confidence
89% confidence
Finding
The skill asks the agent to write persistent configuration into CLAUDE.md, which expands its capability from social posting into local file modification. Although framed as convenience, persistent writes create a trust boundary issue because a skill that can modify local project or global agent context can influence future runs and may store sensitive account metadata in a broadly reused file.

Static analysis

No suspicious patterns detected.