Security audit
Security checks for vulnerabilities and agentic risk
This is a local prompt-compression skill with some rough documentation, but no evidence of hidden access, persistence, exfiltration, or destructive behavior.
Install only if you want local, deterministic prompt/context compression. Review outputs before reusing them, especially for legal, medical, financial, security, or policy text, because any compression can drop nuance. Prefer the documented CLI entry point and ignore the __pycache__ integration suggestion.
The trigger conditions are broad natural-language phrases without firm scope limits, so an agent may invoke this skill in situations where compression is inappropriate. In a prompt-compression skill, over-triggering can cause unintended loss of important context, safety constraints, or task-critical detail, which can indirectly degrade downstream security and reliability.
The description says to use the skill when the user says phrases like '把这段压缩一下还别丢重点' or when they want to lower context cost, which are broad requests that can arise in many ordinary conversations. It does not provide negative examples or explicit constraints to distinguish this skill from general summarization or editing tasks, increasing the chance of unintended invocation.
The documentation tells integrators to import or read from __pycache__, which is not a stable or intended public interface and bypasses the declared CLI entry point. This can lead consumers to execute stale or unintended bytecode, create version-skew bugs, and normalize unsafe loading patterns that are harder to review and validate.
The workflow instructs the agent to 'spot trigger words' such as 'compress' and 'save tokens', which are generic phrases rather than tightly scoped activation commands. Without stronger boundaries or negative examples, this can cause the skill to activate on routine user language not specifically requesting this compressor.
The Chinese documentation repeats the unsafe recommendation to read compiled modules from __pycache__, reinforcing a contradictory and risky integration path. Because __pycache__ contents are implementation artifacts, using them operationally can result in unreviewed code paths, stale bytecode use, and inconsistent behavior across environments.
The trigger description is very broad and includes many common user phrasings about long prompts, summarization, token reduction, and fitting content into context windows. That can cause the skill to activate outside a narrowly bounded use case, leading an agent to compress content when the user did not explicitly request this specific transformation, which may alter meaning or omit safety-critical details.
文件标题和核心说明均以中文撰写，且未见说明该技能是否仅面向中文环境、是否可根据用户语言切换，或是否需要用户明确选择语言。按规则，强制或默认单一语言而无 opt-in/例外说明，可能构成自然语言层面的语言策略问题。
The document is explicitly bilingual, presenting both English and Chinese content as a fixed requirement rather than offering a user-selectable language or locale choice. Under the policy rule, forcing a specific language or locale behavior without opt-in can be a natural-language policy issue, even when multiple languages are included.
All user-facing natural-language guidance in this file is presented only in Chinese, with no indication that the user can choose another language or that the skill is intentionally limited to a Chinese-speaking context. Under the policy, a language constraint without opt-in or justification is a natural-language policy concern.
This file contains natural-language strings such as the module docstring and CLI help text exclusively in Chinese, which effectively forces a specific language for users. The policy allows locale constraints only when users are given a choice or the restriction is clearly documented and justified, neither of which appears here.
