Back to skill

Security audit

Dailytech

Security checks for vulnerabilities and agentic risk

Overview

This tech-news skill is broadly coherent, but its push feature can create persistent scheduled messages to external channels without enough consent, authorization, or lifetime controls.

Install only if you want a bilingual tech briefing and are comfortable with persistent morning/evening pushes. Before enabling push, verify the target userId, channel, timezone, and disable path, and avoid running the toggle for recipients you do not control.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • 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
Findings (2)

T01 · Skill Instruction Hijacking

Note
Location
SKILL.md:65
Finding

Skill instructions impose branded output and fixed artifact behavior

Content
View full analysis
Remediation
View remediation

T06 · System Persistence

Warning
Location
scripts/push-toggle.js:22
Finding

Skill can establish persistent twice-daily scheduled execution

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (11)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description suggests a polished daily AI/tech news product with bilingual visual cards and morning/evening pushes. The supplied code instead performs a local user file lookup and then emits a Chinese instruction prompt to search for and generate an evening tech report. While evening tech news is broadly related, the implementation shown is materially narrower and different: it does not generate a visual card, does not demonstrate 60-second formatting, does not specifically target the declared AI/LLM/startup/dev-tools categories, and accesses local user files despite no declared permissions. Therefore the description does not accurately represent the observed code behavior.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The code is partially aligned in that it prepares a morning AI/tech news briefing prompt for the current date. However, the declared description promises a broader skill: daily 60-second AI/tech news, bilingual visual cards, and both morning and evening push. This code chunk does not generate a visual card, does not implement bilingual output, and does not perform any push delivery. It also reads from local user JSON files, which is an undeclared resource access given the empty permissions declaration. Additionally, the weekday-based vocabulary theme variables appear unrelated to the stated news purpose. Overall, the actual behavior is narrower and materially different from the declared functionality.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The manifest keywords include generic phrases such as "tech news", "AI news", "latest tech", and their Chinese equivalents, which are broad enough to match ordinary user requests unrelated to this specific skill. The file does not provide trigger boundaries, exclusion conditions, or negative examples to clarify when the skill should or should not activate.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill is presented primarily as a news-briefing generator, but it also includes commands to enable recurring outbound push delivery to multiple third-party channels. That additional operational capability expands the security and privacy surface because it can trigger ongoing external communications tied to a user identifier without any documented consent, authorization, or abuse controls.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The push-management section enables recurring notifications to external channels, but it does not warn users that activating it may produce ongoing outbound messages. That omission is risky because recurring delivery can continue beyond the current interaction and may be abused for spam, harassment, or unexpected disclosure through linked messaging accounts.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

Managing push subscriptions and cross-channel delivery is more sensitive than merely generating a bilingual news card because it introduces persistent messaging behavior and handling of user/channel identifiers. Without clear justification, guardrails, and consent language, this can be abused for unwanted notifications or messaging to external platforms.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The generated instruction explicitly requires Chinese phrasing and a bilingual output format ('中英双语') with no user opt-in or configurable locale. This is a natural-language policy concern because the script imposes a language/locale behavior rather than offering a choice or documenting a region-specific need.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The generated instruction requires Chinese output elements such as Chinese titles, summaries, and weekday/date phrasing, and the script also hard-codes Chinese vocabulary themes and weekday names. This is a natural-language locale policy concern because the skill enforces a specific language/locale without offering opt-in or an alternative.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The cron configuration hard-codes tz:'Asia/Shanghai', and the user-facing strings are entirely in Chinese, which imposes a specific locale on all users. The policy allows locale constraints only when users are given a choice or the regional limitation is explicitly justified, neither of which appears in this file.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The description states the output is a "Bilingual EN/CN visual card," which imposes a specific language combination by default. There is no indication that the user can choose language preference or opt out of one locale.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
85% confidence
Finding

The instruction to save output under /mnt/user-data/outputs creates a file on user-accessible storage without explicitly informing the user that local storage will be used. While low severity, undisclosed file creation can surprise users, create retention/privacy concerns, and normalize writing artifacts to disk without clear consent.

Content

No source excerpt is available for this finding.