Back to skill

Security audit

Crypto Encoder Pro

Security checks for vulnerabilities and agentic risk

Overview

This crypto helper is mostly coherent, but it quietly retains sensitive inputs in memory and gives unsafe password-storage guidance.

Review before installing. Avoid using this skill with real credentials, JWTs, passwords, API keys, or bearer tokens unless logging is removed or redacted. Do not follow its MD5/SHA256 password-storage guidance; use a dedicated password-hashing library instead.

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 (2)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:131
Finding

Sensitive Inputs and Encoded Outputs Retained in Cleartext History

Content
View full analysis
str: """Base64 encoding""" result = base64.b64encode(text.encode("utf-8")).decode("ascii") self._log("base64_encode", text, result) return result ``` ```python def url_encode(self, text: str) -> str: """URL encoding""" from urllib.parse import quote result = quote(text, safe="") self._log("url_encode", text, result) return result ``` ```python def sha256(self, text: str) -> str: """SHA256 hash""" result = hashlib.sha256(text.encode("utf-8")).hexdigest() self._log("sha256", text, result) return result ``` ```python def _log(self, operation: str, input_val: str, output_val: str): self.history.append({ "operation": operation, "input": input_val[:50], "output": output_val[:50], "timestamp": datetime.now().isoformat() }) ``` ### Technical Analysis Base64, URL encoding, and SHA-256 operations automatically pass their input and output to `_log`. The logger retains the first 50 characters of each value in the mutable `history` property for the lifetime of the `CryptoEncoder` instance. Inputs to an encoding utility may include Basic Authentication values, bearer tokens, API keys, personal data, or other secrets. Truncation to 50 characters is not effective redaction: many credentials fit entirely within that limit, while prefixes may still contain security-relevant information. This retention is not required to perform the declared transformations and therefore exceeds minimum data-handling requirements. Base64 and URL encoding themselves are necessary for the Skill's declared purpose, and the audit found no network transmission or covert exfiltration mechanism. The risk arises from unnecessary in-process ...[truncated 997 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:70
Finding

Unsafe Password-Storage Guidance Using Fast Hash Functions

Content
View full analysis
str: """MD5 hash""" return hashlib.md5(text.encode("utf-8")).hexdigest() ``` ```markdown ## Usage Scenarios 1. **API Authentication**: Base64-encoded Basic Auth and bearer tokens 2. **Data Integrity**: SHA256 file-hash verification 3. **Password Storage**: MD5/SHA256 hashing (salted) 4. **Token Processing**: Parse JWTs and generate API keys 5. **Web Development**: Encode URL parameters and generate CSRF tokens ``` The implementation also supplies a direct, unsalted SHA-256 operation: ```python def sha256(self, text: str) -> str: """SHA256 hash""" result = hashlib.sha256(text.encode("utf-8")).hexdigest() self._log("sha256", text, result) return result ``` ### Technical Analysis MD5 and SHA-256 are fast general-purpose digest algorithms. They are unsuitable for password storage because attackers can calculate large numbers of candidate hashes per second using commodity GPUs or specialized hardware. The documentation identifies MD5/SHA-256 as a password-storage use case and mentions salting, but the provided methods do not accept or generate salts. They also provide no configurable work factor, memory-hard processing, or password-hashing format. A user following the example may consequently store deterministic, unsalted, fast password digests. MD5 additionally has well-known collision weaknesses. Although collision resistance is not the primary property exploited in password cracking, its use further encourages obsolete cryptographic practices. ### Attack Path 1. A developer follows the Skill's password-storage guidance. 2. The developer hashes user passwords with `md5` or `sha256`. 3. The application stores the resulting fast, potentially unsalted digest in a database ...[truncated 794 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (7)

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The manifest description claims a 'Premium version with full implementation, examples, and enterprise support' but does not describe the advertised crypto/encoding functions of the skill. This mismatch can mislead reviewers and users about the skill’s actual purpose, increasing the risk of deceptive packaging, incorrect trust decisions, or hidden functionality going unnoticed.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation keywords include very generic terms like '编码', 'hash', and 'uuid', which can cause the skill to trigger on ordinary requests that were not intended for this tool. In an agent environment, overbroad invocation increases the chance of accidental handling of sensitive inputs and unexpected transformation or logging of user data.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill advertises JWT parsing without warning that parsing does not verify the signature, which can mislead users into trusting attacker-controlled token contents. In a security-themed tool, omission of this distinction is particularly risky because users may treat decoded claims as validated authentication data.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The manifest description at L005 mentions hash calculation but does not mention keyed signing or message authentication. HMAC-SHA256 is a distinct cryptographic operation from simple encoding/decoding and hashing, so the implemented behavior exceeds the stated description.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest description at L005 limits the skill to Base64/URL encode-decode, SHA256/MD5 hashing, UUID generation, and JWT parsing. The code additionally provides a password strength analysis feature, which is a separate security-auditing capability not disclosed in the manifest description.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The _log function stores portions of raw inputs and outputs in memory, but the skill description does not warn users that potentially sensitive material such as tokens, passwords, secrets, or decoded content may be retained. In this skill context, users are likely to process exactly that kind of sensitive data, so silent retention materially increases privacy and secret-exposure risk.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The manifest description is presented in Chinese, and the rest of the skill content is primarily Chinese with some English terms, but there is no indication that this language choice is optional or region-specific. This can conflict with a language-choice policy when users are not given an opt-in or alternative.

Content

No source excerpt is available for this finding.