Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
ts-prompt-optimizer
v1.0.1冬冬主人定制提示词优化器 - 完全个性化 多模型支持 智能路由集成 使用前缀 "ts:" 触发
⭐ 0· 25·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
Name/description (prompt optimizer, multi-model routing) match the included scripts (optimizer, config manager, config wizard). Requiring python3 and optional DEEPSEEK_API_KEY / BAILIAN_API_KEY is coherent with multi-model support. However the registry metadata in the top-level listing is malformed (shows "[object Object]" for required env vars) and the registry claims 'instruction-only' while the package contains multiple executable scripts — these inconsistencies reduce trust.
Instruction Scope
SKILL.md instructs running the included Python scripts (config_wizard.py, ts-config, quick_setup.py) and to provide model API keys to test connections. That scope is appropriate for the stated purpose. Two concerns: (1) the SKILL.md contains many emoji markers and a pre-scan flagged 'unicode-control-chars' which could be an attempt to obfuscate or influence LLM parsers; (2) the config wizard offers to set environment variables and test model connections (network activity) — expected, but worth explicit attention.
Install Mechanism
There is no download/install-from-URL spec in the package—only local Python scripts. No package manager installs or remote archives referenced. This is a lower-risk install mechanism compared with arbitrary downloads.
Credentials
The only API secrets referenced in SKILL.md are DEEPSEEK_API_KEY and BAILIAN_API_KEY which are proportional to multi-model routing. BUT: (1) the registry metadata is malformed and doesn't correctly declare those env vars, which is suspicious; (2) scripts (config_wizard/config_manager) will read/write config under the user's home (~/.openclaw) and may persist API keys (they call a _set_environment_variable helper and write ts-env-config.json and ts-optimizer-config.yaml). Persisting keys to disk (or injecting into shell profiles) in plaintext is a security/privacy risk — verify exactly how and where keys are stored before providing credentials.
Persistence & Privilege
The skill stores configuration and history under the skill tree and in the user's home (~/.openclaw) and memory/ folders. always:false (not force-included). Autonomous invocation is allowed (platform default). No evidence it modifies other skills or system-wide settings beyond writing its own config, but the config wizard can set environment variables which may persist across sessions — review that behavior if you care about secret handling.
Scan Findings in Context
[unicode-control-chars] unexpected: The SKILL.md and many source files include emoji placeholders and there was a pre-scan hit for unicode-control characters. That is not required for a prompt-optimizer and could be an attempt to obfuscate content or influence LLM parsing. Inspect raw files for hidden/invisible characters and verify the content reviewer/origin.
What to consider before installing
What to check before installing:
- Do not paste high-privilege credentials blindly. Prefer to set API keys yourself in a controlled way (environment variables) and avoid typing them into interactive installers unless you inspect the code first.
- The registry metadata looks broken (shows "[object Object]" for required env vars). Ask the publisher or maintainer to fix the package metadata before trusting it.
- Review the config_wizard and config_manager source (especially any _set_environment_variable implementation) to see whether API keys are written to plaintext files or shell profiles (~/.bashrc, ~/.profile) and where ts-env-config.json/ts-optimizer-config.yaml are saved. If they are stored unencrypted on disk, consider using ephemeral environment variables instead.
- The SKILL.md contains many emoji placeholders and a scan found unicode-control characters. Open the raw files in a hex/hex-aware editor to confirm there are no hidden control characters or obfuscated payloads.
- The skill will perform network requests to test model connections; if you are concerned, run it in a sandboxed environment (VM/container) or with a network monitor to observe outbound endpoints. Expected endpoints: DeepSeek, Bailian (Aliyun console) and optionally other model providers; if you see unknown remote endpoints, do not proceed.
- Because this package includes executable scripts (not just prose), consider running linting and a quick static scan (grep for suspicious functions: subprocess, os.system, requests.post to unexpected domains, writing to /etc, reading ~/.ssh, etc.) before use.
If you want, I can: (a) extract and search the omitted/remaining source files for functions that persist env vars or make network calls, (b) show the _set_environment_variable implementation if present, or (c) point out exact lines that write keys to disk — provide the omitted files and I will analyze them.Like a lobster shell, security has layers — review code before you run it.
ai-assistantvk97av56ka3f95g4ydj9cq8091x846bj8latestvk97fw2tchgs3tj3mtas7enc6zs846pvwproductivityvk97av56ka3f95g4ydj9cq8091x846bj8prompt-optimizationvk97av56ka3f95g4ydj9cq8091x846bj8
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
[TARGET] Clawdis
Binspython3
Env[object Object], [object Object]
