Security audit
Security checks for vulnerabilities and agentic risk
The skill mostly fits a GitHub issue/PR archive-search purpose, but its broad triggers, write-capable metadata, local caching, and GitHub credential flows need review before use.
Review this before installing if you do not want an agent using command-line tools, GitHub authentication, or persistent local caches. Use it only for explicit GitHub issue/PR archive searches on repositories you choose, and avoid letting its broad trigger language apply to unrelated coding, analytics, deployment, or file-processing tasks.
The trigger condition says to use the skill for data analysis, report generation, statistical insights, and visualization—far broader than repository archive search. In a tool-enabled agent, overly broad triggers are risky because they can cause unintended invocation of a skill with exec/read/write capabilities in workflows that never needed repository access or command execution.
The top-level description says to use the skill for code generation, programming assistance, debugging, testing, deployment, and other broad workflows, even though the skill is supposed to search repository issue/PR archives. This mismatch can make an agent over-select the skill in unrelated contexts, leading to execution of shell commands or other side effects when a simpler, safer tool would suffice.
The description explicitly states '支持中文交互', and the document is written as though Chinese interaction is the expected mode, but it does not offer the user a language choice or explain a necessary locale constraint. This can violate language/locale policy when users are not given an opt-in or alternative.
The skill describes creating local archives, syncing data, and export-like operations, but does not clearly warn users about persistence, storage location, retention, overwrite behavior, or network/data effects. In context, the skill also declares write and exec tools, so hidden or poorly explained persistence increases the risk of unintentionally storing repository data locally or mutating existing cache state.
The skill states the free version does not support real-time sync, but later instructs users to run sync/update and even live-style validation flows. Contradictory instructions are dangerous in an exec-enabled skill because they create ambiguity about when networked or state-changing commands are allowed, making unintended external access or local state mutation more likely.
The manifest presents the skill as a narrow repository issue/PR archive search tool, but the later '核心功能' section broadens it into generic automation, file processing, API integration, and command execution. This scope drift can cause an agent or user to invoke the skill for tasks far outside its declared purpose, increasing the chance of unnecessary file writes, external calls, or shell execution under an apparently low-risk label.
The capability table says multi-repository management is not supported, and the limitations repeat that claim. However, the FAQ suggests using 'gitcrawl list-repos', which implies the skill can manage or at least enumerate multiple repositories, contradicting the stated limitation.
No suspicious patterns detected.
