Smart Expense Tracker Cn
v1.2.0智能记账助手 v2.1 | AI Expense Tracker. 支持语音记账、OCR发票识别、预算提醒、周月报推送、智能分类、多账本管理、数据导出。触发词:记账、花了、消费、支出。
⭐ 0· 390·3 current·4 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The skill claims OCR and voice capabilities; requiring tesseract (OCR) and jq (JSON handling) would be reasonable for that purpose. However, the registry metadata you provided lists no required binaries, while the embedded SKILL.md metadata lists requires: [jq, tesseract]. This mismatch is a configuration inconsistency that should be clarified.
Instruction Scope
The SKILL.md is user-facing and does not instruct the agent to read unrelated system files or environment variables, which is good. However, the pre-scan found 'unicode-control-chars' in the SKILL.md — hidden control characters can be used to hide or alter prompts and thus manipulate agent behavior. Also the document references sending reports to email, QQ/WeChat, and an API for Enterprise tier but provides no technical details or declared credentials; that ambiguity could mask later steps that require sensitive credentials or external endpoints.
Install Mechanism
This is an instruction-only skill with no install spec or code files, so nothing is downloaded or written by an installer. That lowers install-time risk. The only install-related concern is the SKILL.md's own metadata listing required binaries (jq, tesseract) which are expected for OCR/JSON processing but are not enforced by the registry metadata.
Credentials
The skill declares no required environment variables or primary credential, which is appropriate for a local instruction-only helper. That said, features mentioned (email sending, QQ/WeChat integration, API for Enterprise) normally require credentials or tokens — none are declared. This omission is a gap: if connectors are implemented later, they should be explicitly declared and justified.
Persistence & Privilege
The skill does not request always: true and is user-invocable with normal autonomous invocation allowed. It does not request system-wide config paths or persistence in its manifest, which is proportionate for this type of skill.
Scan Findings in Context
[unicode-control-chars] unexpected: Hidden Unicode control characters in an instruction file are not needed for a user-facing guide and can be used to manipulate LLM parsing or hide additional instructions. This is suspicious and should be explained or the file should be provided without such characters.
What to consider before installing
What to consider before installing:
- Clarify the binary requirements: SKILL.md lists jq and tesseract (reasonable for OCR + JSON), but the registry metadata you provided lists none — ask the publisher to confirm required binaries and versions.
- Ask for the skill source or a repository / homepage. Unknown source increases risk.
- Ask the author to provide a plain-text SKILL.md with control characters removed and to explain why unicode control characters were present. Use a hex or control-character viewer to inspect the file yourself.
- Confirm how external integrations work: the doc mentions emailing reports, QQ/WeChat bots, and an API for enterprise — request details about endpoints, where images/receipts are uploaded, and what credentials are needed. Do not supply email/QQ/API credentials until you understand the data flow and storage policies.
- Request a privacy/security statement: where user data (images, receipts) is stored, whether OCR is performed locally (tesseract) or sent to remote services, retention policy, and whether data is shared with third parties.
- If you plan to test it, run it in a restricted sandbox account without real personal or financial data and without privileged credentials.
- If the publisher cannot explain the control characters or provide transparent integration details, avoid installing or granting credentials.
In summary: the skill's declared capabilities are plausible, but the hidden-control-character finding, metadata mismatch, and vagueness about external connectors are red flags — get clarification and the source code before trusting it with real data or credentials.Like a lobster shell, security has layers — review code before you run it.
latestvk977538eqv14ehrtepp8g762e982b2yz
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
