Back to skill

Security audit

日历·催化剂

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report weaknesses in its calendar checks and broader workflow concerns involving external queries, credential handling, and required billing steps.

Consider not installing until you are comfortable with the reported date-verification gaps, the external-query privacy concerns, and the API-key instructions. Skillspector coverage was partial: 106 of 107 components were scanned, and one static analysis was degraded, so the evidence is incomplete.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (11)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

Calendar validation is a legitimate supporting implementation detail, so the absence of calendar generation or reminders in this individual script is not itself a mismatch. However, the implemented acceptance rules materially undercut the description's explicit guarantees. Matching a source reference or finding an anchor somewhere in evidence does not establish that the calendar date is disclosed or that statutory applicability conditions hold. The filtering also contains exceptions beyond the two expressly allowed. Consequently, PASS does not establish the strict evidence and neutrality contract advertised by the description. No unrelated execution, external transmission, or resource access is apparent.

Content

No source excerpt is available for this finding.

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · SKILL.md (reported line 243)May include surrounding context.

md
echo "=== 检查1 runner ===" && [ -f "$HOME/.cue/cue-skills/cue-research/scripts/research_run.py" ] && echo "已安装" || echo "未安装!"
echo "=== 检查2 API Key ===" && [ -n "$CUE_KEY" ] && echo "已配置" || echo "未配置!"
echo "=== 检查3 Cue 服务 ===" && curl -sS --max-time 10 "https://cuecue.cn/api/health" -H "Authorization: Bearer $CUE_KEY"
echo "=== 检查4 剩余积分 ===" && curl -sS --max-time 10 "https://cuecue.cn/api/v1/billing/subscriptions/current" -H "Authorization: Bearer $CUE_KEY" | python3 -c "
import sys,json
d=json.load(sys.stdin) or {}
d=d.get('data',d) or {}

A successful check does not require evidence establishing the event's stated date.

High
Category
Excessive Agency
Confidence
99% confidence
Finding

Evidence checking is optional, and even when enabled it checks anchor presence rather than whether the cited passage establishes the event and its stated date. A fabricated date can therefore pass alongside a genuine announcement identifier, violating the calendar's central provenance guarantee and potentially misleading investors or triggering reminders on incorrect dates; the documented human-review limitation does not make PASS proof of date validity.

Content

No source excerpt is available for this finding.

The workflow does not warn that holdings-derived queries are sent to external Cue services.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The workflow uses holdings-derived issuer lists to query external Cue services without explicitly explaining that this discloses portfolio-related interests to the provider. Naming Cue and obtaining approval for spending do not establish informed consent to that data transfer; individually public issuer names can reveal sensitive information when associated with a user's holdings.

Content

No source excerpt is available for this finding.

The workflow omits a privacy warning about sending holdings-derived queries to Cue.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The documented workflow derives external data requests from a user's holdings without warning that Cue receives those issuer selections. Although the underlying announcements are public, the association between a user and a set of investments can be private, and consent to charges does not address that disclosure.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

API-key onboarding requests a secret in chat without warning about credential exposure.

Medium
Category
Output Handling
Confidence
99% confidence
Finding

The onboarding instructions explicitly ask the user to paste an API key into chat before it is written to local configuration. This exposes a reusable credential to conversation storage and potentially model or logging systems, enabling unauthorized API usage or credit consumption if the transcript is accessed.

Content

No source excerpt is available for this finding.

The research workflow lacks a privacy warning that portfolio-related queries are sent to Cue.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The research command sends issuer identifiers and selected events to Cue, potentially revealing holdings-related interests under an authenticated account. The workflow obtains spending authorization but does not explicitly disclose this privacy consequence, including when queries run later through scheduled reminders.

Content

No source excerpt is available for this finding.

Credential onboarding instructs users to disclose an API key in chat while claiming the agent has no right to view it.

Medium
Category
Data Exfiltration
Confidence
99% confidence
Finding

The text claims the agent has no right to view the API key while requiring the user to send that same key to the agent in chat. This contradictory assurance can encourage credential disclosure under a mistaken expectation of secrecy, exposing the account to unauthorized use if conversation records leak.

Content

No source excerpt is available for this finding.

The workflow mandates billing checks and promotional prompts even for calendar-only delivery.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The skill mandates an authenticated billing lookup and promotional interaction after calendar-only delivery, even though that task is presented as free and independent of research onboarding. This unnecessarily expands the task into credential use and account access, while rigid promotional instructions displace the user's requested workflow.

Content

No source excerpt is available for this finding.

The checker permits additional rating and sentiment exemptions beyond the manifest's two allowed exemptions.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The checker introduces exemptions beyond the manifest's two permitted exceptions: it removes book-title-bracketed text before sentiment scanning, skips headings, and exempts entire non-table lines from rating checks when they contain a prohibition marker. An attacker could place prohibited recommendations in a heading or append an unrelated prohibition to a recommendation and evade detection, undermining this financial calendar's explicit no-ratings/no-sentiment guarantee.

Content

No source excerpt is available for this finding.