Security audit

skill-upgrader

Security checks for vulnerabilities and agentic risk

Overview

This is a documented skill-improvement helper with broad but disclosed authority to revise one chosen skill, and I found no hidden code, credential access, or background behavior.

Install this only if you want an agent to help revise OpenClaw skill instructions. Before accepting an upgraded skill, review the diff, confirm it only changed the intended skill, and narrow any generated trigger rules that could activate on ordinary unrelated requests.

SkillSpector

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The example trigger phrase is broad enough to overlap with ordinary user requests about writing, which can cause the meta-skill to activate when the user did not explicitly ask to modify or upgrade a skill. In a meta-skill that rewrites other skills, this increases the chance of unintended instruction interception, prompt collisions, and behavior changes across unrelated tasks.

Vague Triggers

High

Confidence: 97% confidence
Finding: The description repeatedly uses universal activation terms like 'ANY existing skill' and positions the skill as applicable whenever a user needs to 'enhance, refactor, or evolve any skill.' In an agentic system, this makes triggering criteria overly broad and can cause the meta-skill to intercept ordinary requests, expanding its authority and increasing the chance of unintended modification of unrelated skills or workflows.

Vague Triggers

High

Confidence: 96% confidence
Finding: The quick-start says to execute the upgrade protocol when asked to 'upgrade, enhance, or improve ANY skill,' which is ambiguous and broad enough to match many routine prompts. Because this is a meta-skill that rewrites or restructures other skills, ambiguous invocation materially raises the risk of accidental takeover, unrequested transformations, or cascading changes across the skill ecosystem.

Vague Triggers

Medium

Confidence: 92% confidence
Finding: The guidance 'Activate when: any writing task' is an overbroad trigger that can cause the skill to intercept a very large class of requests without sufficient scoping or anti-trigger constraints. In a meta-skill designed to modify or enhance other skills, this increases the chance of unintended activation, instruction interference, and inappropriate takeover of tasks better handled by more narrowly scoped skills.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: Using generic phrases like 'Write a...', 'Create a...', or 'Draft a...' as explicit triggers is too broad because these phrases occur in routine user requests that may not warrant this skill's involvement. In the context of a powerful 'skill-upgrader' meta-skill, such broad triggers can lead to excessive auto-activation, context hijacking, and accidental application of upgrade logic to unrelated tasks.

Static analysis

No suspicious patterns detected.