Security audit
Security checks for vulnerabilities and agentic risk
This is a documentation-only skill for fetching Bing's daily wallpaper, with a real but low-impact routing-quality issue from unrelated trigger keywords.
Before installing, be aware that this skill may be invoked for unrelated image tasks because its keywords are too broad. It should be narrowed to Bing daily wallpaper terms, but the inspected artifact does not show malicious code, persistence, local data access, or credential misuse.
The trigger phrases are overly broad and include unrelated image-processing intents, which can cause this skill to activate for tasks outside its actual functionality. In an agent ecosystem, broad mismatched triggers are dangerous because they can hijack routing, leading to wrong tool selection, failed operations, or unexpected data being sent to an irrelevant endpoint.
The common-keywords section repeats unrelated keywords that do not match the declared single-endpoint scope. Because keyword-based activation often drives automatic tool choice, these ambiguous triggers materially increase the chance of incorrect invocation and unreliable behavior.
The English trigger phrases are overly broad and include functions unrelated to this endpoint, which increases the chance of unintended invocation. In an agentic environment, broad trigger poisoning can misroute user requests to an incorrect skill, causing incorrect outputs, wasted calls, and potential chaining into unsafe or unauthorized actions if later steps assume the selected skill is appropriate.
The manifest says the skill is for a single Bing daily image endpoint, but the description also advertises unrelated capabilities like image compression and base64 conversion. This mismatch can cause the orchestrator or user to invoke the skill for tasks it does not actually implement, leading to misrouting and unintended API usage.
The inline documentation explicitly states the skill wraps only GET /image/bing-daily, yet elsewhere it lists unrelated image operations. These contradictions weaken safe routing guarantees and can make an agent select this skill for unsupported actions, producing incorrect behavior or confusion in multi-skill environments.
The document is written as Chinese-only operational guidance and presents invocation examples primarily in Chinese, without stating that the user may choose their preferred language or locale. Under the policy, language constraints should be optional, user-selectable, or clearly justified as region-specific.
The document states the skill only wraps GET /image/bing-daily, but its keyword index also advertises unrelated capabilities such as image compression, base64 conversion, SVG conversion, and NSFW detection. This inconsistency can cause the agent to invoke the wrong skill for unrelated prompts, leading to confused task routing and unintended API usage.
The file’s user-facing content is entirely in Chinese, including the title and operational guidance, with no indication that users can choose another language or that the skill is intentionally limited to a Chinese-speaking context. This can violate a language/locale policy when a skill imposes a specific language without opt-in.
No suspicious patterns detected.
