Back to skill
Skillv1.0.0

ClawScan security

张氏财报分析实战操作手册-零会计基础极速排雷 · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignApr 15, 2026, 2:05 AM
Verdict
benign
Confidence
high
Model
gpt-5-mini
Summary
This is an instruction-only financial-analysis manual that is internally consistent with its stated purpose: it asks for no credentials, installs nothing, and instructs only how to extract/report specific financial items and compute metrics.
Guidance
This skill is essentially a step-by-step manual for reading financial statements and computing diagnostic ratios — it appears coherent and low-risk. Before installing or running it: 1) confirm how '自动获取公开数据' is implemented (which data sources/APIs will be accessed and whether any credentials are required); 2) avoid pasting non-public/sensitive documents into the skill (notes can contain PII or confidential counterparty details); 3) if you want to restrict network access or prefer to supply data yourself, provide the financial_data field rather than letting the agent fetch; 4) treat its outputs as heuristics, not definitive legal/financial advice — validate important conclusions with a qualified professional.

Review Dimensions

Purpose & Capability
okName/description (financial statement analysis, fraud/risk screening) align with the instructions: all required steps are extracting line items from financial statements and computing ratios. No unrelated binaries, env vars, or installs are requested. One minor note: the input_schema says '未提供则自动获取公开数据' (will automatically fetch public data if user doesn't provide it) but SKILL.md does not enumerate data sources or required credentials—this is plausible but should be documented.
Instruction Scope
okSKILL.md confines the agent to extracting named fields from income statement, balance sheet, cash flows, and notes, and performing defined calculations and heuristics. It does not instruct reading unrelated system files or secrets. A possible scope consideration: some checks reference report notes and '追溯业务经手人' (tracing business handlers), which may require parsing disclosures that include names/identifiers—still within financial-report scope but may surface PII if present in notes.
Install Mechanism
okNo install spec and no code files — this is instruction-only and therefore low risk from an installation perspective (nothing will be written to disk by the skill itself).
Credentials
noteThe skill declares no required environment variables, credentials, or config paths, which is appropriate for a manual analysis guide. Caveat: the input_schema's promise to 'automatically fetch public data' could require network access or external APIs in practice; no sources or API keys are declared. If the agent will fetch from paid/third-party data providers, credentials would be needed but are not requested—ask the skill author how automatic fetching is implemented or whether the agent will prompt for data sources.
Persistence & Privilege
okThe skill does not request persistent presence (always:false), does not alter other skills or system settings, and has no install-time persistence. Default autonomous invocation is permitted (platform default) but not combined with other concerning privileges.