Security audit
Security checks for vulnerabilities and agentic risk
This skill is a context-budgeting helper with documentation quality issues, but I found no hidden persistence, credential handling, network execution, or destructive behavior.
Install only if you want a context-budgeting and context-audit helper. Expect Chinese-heavy output and some stale or inaccurate usage examples; use the actual CLI help rather than the unsupported documented flags. Do not let it automatically rewrite prompts, memory, or retrieval settings without reviewing the proposed changes.
The changelog states 'Bilingual rewrite (full EN+ZH, English first)', which imposes a language/ordering policy in the skill's natural-language description rather than offering a user choice. This can violate locale/language policy expectations when no opt-in or justification is provided.
The documentation advertises CLI modes and parameters such as --calculate, --budget_limit, and --model_name that are not declared in the interface schema. This creates interface ambiguity and can cause downstream agents or orchestrators to invoke undocumented behavior, weakening validation and opening the door to unsafe assumptions about how the skill should be executed.
The skill claims users can read in-file helper libraries for memory-decay and retrieval-recall algorithms, but the declared entry point and capabilities only describe a budget calculator CLI. This discrepancy broadens the apparent functional scope of the skill beyond its declared contract, which can mislead agents into exposing or depending on undeclared logic and increase the risk of unintended execution paths.
The Chinese usage section repeats undocumented parameters and audit behavior that are absent from the formal interface schema. In multilingual skills, inconsistent declarations are especially risky because one language version may induce behavior that bypasses validation, review, or wrapper constraints assumed by tooling.
The natural-language content is entirely in Chinese and directs agent behavior in that language, but the file does not offer user opt-in or indicate that the skill is intentionally limited to a Chinese-speaking context. Under the stated policy, a language constraint without choice or justification is a locale-policy issue.
The activation text uses very broad, complaint-style triggers such as 'AI越来越笨' and '答非所问', which can match ordinary user frustration rather than a clear request to invoke this specific skill. That increases the chance of unintended activation, causing the agent to steer into this skill's workflow when the user's actual intent is different.
This file contains natural-language instructions and status messages that force a specific language/locale for users. Under the policy, a skill should not impose a language without opt-in unless the locale restriction is clearly documented and justified.
This markdown file includes a substantial policy and licensing section written primarily in Chinese, including imperative language such as '务必并读' and a required unchanged disclosure block. That can function as a language/locale constraint for users who do not read Chinese, and the file does not offer an explicit language-choice mechanism for this section.
Dynamic getattr() with a non-literal attribute name can access arbitrary object attributes, potentially bypassing access controls.
blocks[args.block] = read_block("-")
else:
for b in ("system", "memory", "retrieval", "tool", "history"):
v = getattr(args, b)
if v != "":
blocks[b] = read_block(v)