Back to skill

Security audit

yoooclaw-daily-morning-brief-en

Security checks for vulnerabilities and agentic risk

Overview

The skill has a coherent morning-brief purpose, but it should be reviewed because it can broadly query sensitive work and personal notifications and runs a local command without clear consent, scoping, or input-safety controls.

Install only if you are comfortable with the skill querying and summarizing recent notifications across connected work and personal sources. Prefer limiting connected apps and watched groups, and avoid entering free-form time ranges unless the platform validates them safely.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:24
Finding

Command Injection Through Unvalidated Time-Range Arguments

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 24-31
Vulnerability Type: OS command injection through user-controlled command arguments
Risk Level: High

Vulnerable Code

text
1. To query notifications within a specified time range, use exec to execute:
If the user does not provide the corresponding time, query notifications from yesterday to the current time:
```text
command: openclaw ntf search --from start time --to end time
yieldMs: 30000

Example:

text
command: openclaw ntf search --from 2026-03-01T00:00:00+08:00 --to 2026-03-09T23:59:59+08:00

Technical Analysis

The skill instructs the agent to execute openclaw ntf search using a time range that may originate from the user's request. It does not require strict ISO-8601 validation, define an allowlist of accepted characters, or require a shell-free argument-array execution interface.

If the execution tool passes the constructed command through a command shell, shell metacharacters embedded in a malicious --from or --to value may be interpreted as command separators, substitutions, redirections, or pipelines rather than as part of the timestamp. This can permit arbitrary command execution under the identity and privileges of the agent process.

Exploitability depends on the execution tool accepting a command string through a shell and on the agent inserting user-supplied values without safe encoding. The skill provides no controls that prevent this unsafe execution path.

Attack Path

  1. An attacker asks for a morning report over a deliberately crafted time range.
  2. The attacker places shell syntax in the supplied start or end time.
  3. Following the skill workflow, the agent inserts the supplied value into the openclaw ntf search --from ... --to ... command.
  4. A shell-backed exec implementation interprets the injected syntax.
  5. The attacker's additional command executes with the permissions of the ...[truncated 796 chars]
Remediation
View remediation

Remediation Suggestions

  1. Parse both time values with a strict ISO-8601 date-time parser before command execution. Reject invalid dates, trailing content, control characters, whitespace beyond the defined format, and shell metacharacters.
  2. Invoke the executable through a shell-free API using a fixed argument array, for example:
    text
    ["openclaw", "ntf", "search", "--from", validatedStart, "--to", validatedEnd]
    
  3. Do not concatenate user-controlled values into a command string. If the platform only supports string commands, apply robust platform-specific escaping and reject any input outside a narrowly defined timestamp allowlist.
  4. Generate omitted time values internally rather than accepting free-form replacements.
  5. Run the command with least privilege and restrict its filesystem, credential, and network access through sandboxing.
  6. Require confirmation or reject the request when a supplied time value cannot be normalized unambiguously.
  7. Add tests covering command separators, command substitution, quoting characters, redirection operators, newlines, and malformed timestamps.
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill is explicitly named with an '-en' suffix, and the output template is written as fixed English output, but the file does not state that the user chose English or can select another language. This creates a language-policy concern because it appears to enforce a locale without opt-in or justification.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger phrases are broad enough to match common requests such as 'Help me organize the morning report' or 'What important information and to-dos are there today,' which can cause the skill to activate in situations where the user did not intend broad notification retrieval and summarization. In this context, over-broad activation is risky because the skill may pull and process sensitive messages from multiple integrated sources based on an ambiguous prompt.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill processes notifications from a wide range of work and personal sources, including private chats, email, calendar, and family-related messages, but it does not provide a clear user-facing warning about the sensitivity of the data being collected and summarized. This omission undermines informed consent and raises the risk of unexpected aggregation of highly sensitive information into a single output, increasing privacy and confidentiality exposure.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill instructs the agent to execute a local CLI command (openclaw ntf search) to retrieve notification data. Even though this appears functionally related to the skill, it expands the skill from pure summarization into local data access and can expose sensitive personal and work notifications across multiple sources without an explicit per-use consent or clear scoping safeguards. The context makes this more dangerous because the queried data includes mixed personal, corporate, and family communications.

Content

No source excerpt is available for this finding.