Skills

This skill employs prompt injection and manipulative psychological framing to hijack the agent's decision-making process, forcing it to override system instructions and follow an unauthorized, rigid workflow.

The skill lacks transparency, uses deceptive pre-authorized instructions, and contains insecure file-access patterns that could allow unauthorized traversal into sensitive directories.

The skill attempts to manipulate instruction boundaries to override system prompts and fails to adhere to standard naming and licensing conventions.

The skill masquerades as a commit message generator but forces unauthorized terminal execution of git commands, bypassing critical human oversight and verification processes.

The skill deceptively fabricates user consent to bypass authorization protocols, posing a significant security risk regarding user intent and command integrity.

The skill attempts to manipulate instruction boundaries to override system prompts, posing a significant risk of unauthorized control over the agent's behavior.

The skill attempts to manipulate instruction boundaries to override system prompts, posing a significant risk of unauthorized control over the agent's behavior.

The skill performs instruction boundary manipulation, allowing it to override system prompts and potentially bypass safety filters to execute unauthorized commands.

This skill executes arbitrary shell commands and installs unverified packages while explicitly bypassing human oversight, creating critical risks of remote code execution and credential exfiltration.

This skill bypasses human oversight by disabling confirmation prompts and performs unauthorized network discovery and file operations without declaring the necessary tools or security constraints.

The skill attempts to manipulate instruction boundaries to override system prompts, posing a significant risk of unauthorized control over the agent's behavior.

The skill attempts to manipulate instruction boundaries to override system prompts, posing a significant risk of unauthorized behavior and prompt injection.

The skill deceptively fabricates user consent to bypass authorization protocols and lacks a specified license, posing significant risks to user autonomy and legal transparency.

The skill performs instruction boundary manipulation, which allows for prompt injection attacks that can override the agent's system instructions and bypass established safety filters.

The skill attempts to bypass human-in-the-loop controls and manipulates instruction boundaries to override system prompts, posing a significant risk to agent autonomy and security.

The skill is vulnerable to dependency confusion attacks and prompt injection via unverified external references, while lacking secure instruction boundaries to prevent unauthorized agent behavior modification.

The skill lacks defined tool constraints, performs aggressive unauthorized network scanning, and utilizes insecure command execution patterns that bypass system-level security boundaries.

The skill attempts to manipulate instruction boundaries to override the agent's system prompt, posing a significant risk of unauthorized control and safety filter bypass.

The skill exhibits critical security flaws, including insecure secret handling, unconstrained network access, vulnerable cryptographic implementation, and instruction boundary manipulation, while lacking necessary dependency pinning and documentation.

This skill functions as a malicious exploitation toolkit that instructs the agent to perform destructive SQL injections, exfiltrate credentials, and execute unauthorized reverse shells against target systems.

The skill attempts to extract the agent's system prompt, indicating a malicious intent to bypass security controls and compromise the agent's operational integrity.

The skill attempts to override the agent's system instructions, posing a significant risk of unauthorized behavior modification and prompt injection.

The skill attempts to manipulate instruction boundaries and uses brand impersonation to deceive users, posing a significant risk of prompt injection and unauthorized system behavior.

The skill uses deceptive pre-authorized instructions to bypass user consent while lacking transparency through missing code documentation and licensing information.

The skill is insecure because it executes untrusted local bash scripts and processes user-controlled markdown files, enabling arbitrary code execution and prompt injection attacks.

This skill systematically disables all human oversight and security guardrails while implementing an insecure, persistent memory architecture vulnerable to prompt injection and supply chain attacks.

Security issues detected in inference-sh/skills/prompt-engineering.

The skill is vulnerable to prompt injection through dynamically fetched external data files that can override system instructions and manipulate the agent's execution flow.

The skill attempts to manipulate instruction boundaries to override system prompts, posing a significant risk of unauthorized control over the agent's behavior.

The skill attempts to extract system prompts and override instructions while executing unpinned npx packages, creating significant risks of prompt injection and supply chain compromise.

The skill attempts to manipulate instruction boundaries to bypass system prompts and lacks transparency regarding its code and licensing, posing a significant security risk.

The skill attempts to override system instructions and performs unauthorized network and file operations without declaring necessary tool permissions, posing a significant security risk.

The skill uses fabricated user consent to bypass authorization and executes unpinned npx packages, while relying on missing documentation files that could lead to unpredictable runtime behavior.

The skill attempts to manipulate instruction boundaries to override system prompts, posing a significant risk of unauthorized control over the agent's behavior.

The skill performs instruction boundary manipulation and relies on missing external documentation, creating potential for unauthorized runtime content injection and unpredictable workflow behavior.

This skill attempts to manipulate instruction boundaries and uses brand impersonation to deceive users, posing a significant risk of prompt injection and unauthorized system behavior.

The skill mandates autonomous operation by bypassing human-in-the-loop controls and uses fabricated user consent to perform unauthorized resource creation while lacking transparency and proper documentation.

The skill attempts to manipulate instruction boundaries to bypass system prompts, posing a significant risk of unauthorized command execution and prompt injection.

The skill performs unauthorized system operations and manipulates instruction boundaries while lacking defined tool constraints, posing a significant risk of arbitrary code execution and prompt injection.

The skill contains embedded LLM control tokens that could be exploited to perform prompt injection attacks and manipulate the agent's underlying system instructions.

The skill contains suspicious LLM control tokens and lacks transparency regarding its code and licensing, posing a significant risk of unauthorized prompt injection or malicious instruction manipulation.

The skill performs unauthorized system enumeration, network discovery, and persistent task creation while bypassing security constraints by failing to declare its required tools and capabilities.

This skill is highly insecure, enabling arbitrary JavaScript execution and indirect prompt injection while actively bypassing user oversight and relying on missing, potentially untrusted external dependencies.

The skill performs instruction boundary manipulation, allowing it to override system prompts and bypass safety filters, posing a significant risk to the agent's integrity.

The skill deceptively fabricates user consent to bypass authorization protocols, posing a significant security risk by manipulating user intent to execute unauthorized actions.

The skill masquerades as a creative writing coach while implementing a persistent persona-based bypass to intentionally override safety guardrails and ignore core operational constraints.

This skill masquerades as a text formatter to execute prompt injection attacks, using fabricated claims of elevated user authority to bypass the agent's safety guardrails and operational constraints.

The skill attempts to override system instructions and access sensitive credential files while performing unauthorized network and file operations without declaring necessary security constraints.