auto-context
v1.1.0智能上下文卫生检查器。分析当前会话的上下文污染程度 （长对话、主题漂移、噪声累积），建议：continue、/fork、/btw 或新会话。 支持手动触发（/auto-context）和自动触发（响应层实现）。 基于 ArXiv 论文和认知心理学研究的多维度评估体系。
⭐ 0· 100·0 current·0 all-time
byJialin@0xcjl
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (context hygiene checker) align with the SKILL.md and the files: the skill only needs access to the current conversation and makes recommendations (/fork, /btw, continue, new session). No unrelated environment variables, binaries, or remote endpoints are requested.
Instruction Scope
Runtime instructions focus on analyzing session history (turn counts, semantic similarity/ topic drift, short user inputs, tool-call repetition, compression count). That requires access to the conversation history (expected). The SKILL.md also describes an automatic trigger implemented in the response layer — enabling that requires platform integration and access to responses; verify the platform integration does not grant broader permissions than intended. The SKILL.md contains many duplicated/redundant lines and formatting issues (sloppy but not malicious).
Install Mechanism
Instruction-only skill with no install spec and no code files to execute. package.json and README files are present but no install/download steps or third-party installs are required. This is low-risk from installation perspective.
Credentials
The skill declares no required environment variables, credentials, or config paths. The actions it describes (reading session context and computing metrics) are proportionate to its purpose.
Persistence & Privilege
Flags: always=false and default autonomous-invocation allowed. Autonomous invocation is the platform default; this skill does not request forced/global presence. If you enable the automatic response-layer trigger you will be giving the integration access to examine responses and inject short reminders — confirm that integration is scoped only to the conversation data and not to system credentials or other skills' configs.
Assessment
This skill appears coherent and low-risk: it only analyzes the conversation and produces short suggestions. Before enabling auto-triggering, confirm how the skill will be wired into your agent's response layer (what exact data it will read and whether any extra permissions are required). Because the source is 'unknown' and the SKILL.md has duplicated/poorly formatted sections, consider reviewing the referenced GitHub repo (or the files packaged) yourself or testing the skill in a restricted environment first. If you only want manual use, keep auto-trigger disabled so it cannot post reminders automatically.Like a lobster shell, security has layers — review code before you run it.
latestvk97cd91p7hqjy6aayk317n296n84jz8q
License
MIT-0
Free to use, modify, and redistribute. No attribution required.