Security audit
Security checks for vulnerabilities and agentic risk
ClawScan marked this skill suspicious because scanners report unrelated external catalog requests and a parsing workflow that may send user-provided URLs or paths to an external service without a privacy warning.
Consider the suspicious verdict before installing. Check whether the external catalog and parsing requests are necessary, what information they send, and whether the service is trusted; avoid submitting private links, credential-bearing URLs, or local paths until those details are clear. The Skillspector scan was partial because one referenced path was missing, so the evidence is incomplete.
The supplied command accesses an unrelated catalog endpoint rather than retrieving or parsing shareholder announcements, generating source-linked cards, or identifying missing information. Nothing in the supplied chunk establishes this network request as a supporting implementation detail of the declared workflow. No data upload or other exfiltration is evident.
The supplied command accesses an MCP-catalog endpoint rather than shareholder announcements. It performs no visible announcement parsing, eight-card extraction, source anchoring, or missing-data reporting. Nothing in the supplied chunk establishes this catalog request as a supporting implementation detail of the declared purpose.
The supplied command retrieves and truncates an external catalog rather than performing the declared shareholder-announcement analysis. Nothing in the code establishes that this request supports the declared workflow. No local-data transmission or other exfiltration is shown.
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.
curl -s https://cuecue.cn/api/mcp-catalog | head -c 500
curl -s -X POST "$MCP_ENDPOINT" -H 'Content-Type: application/json'
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
The workflow directs the agent to submit user-provided URLs or local paths to an external parser without requiring informed approval or screening for sensitive inputs. Although shareholder filings are usually public, private URLs, embedded access tokens, and local path metadata can be disclosed; the supplied text does not establish whether local file contents are also uploaded.
Without declared permissions the skill's intent is opaque and cannot be validated.
The instructions route announcement URLs or local paths to an external parsing channel without a privacy warning or an approval gate. Public filings reduce the typical sensitivity, but private links, URL credentials, and filesystem metadata can still cross the trust boundary without informed authorization; actual file-content upload is not established here.
The module documentation at L038–L041 explicitly requires both dividend standard and total dividend in the dividend item, and both repurchase scale and price ceiling in the repurchase item; it states that one candidate cannot satisfy the entire group. However, NUM_CELLS at L089–L123 splits these into four separate items with one required element each, and L416 reports gaps only for those configured elements. Consequently, cards can pass without satisfying the documented paired-element requirements, giving users inaccurate assurance about completeness.