Back to skill

Security audit

APM 메시지 푸시 API

Security checks for vulnerabilities and agentic risk

Overview

This skill is coherent API documentation, but it includes write-capable APM actions and publishes reusable signing salts with weak MD5 signing, so it should be reviewed before installation.

Install only if you trust the APM API owner and can provide a narrowly scoped APM_USER_TOKEN. Treat the token and all generated request signatures as sensitive, avoid logging them, and require explicit user approval before marking messages read or uploading custom information. The publisher should replace the public MD5-plus-salt signing scheme, remove reusable salts from docs if they are intended to be secret, and add write-action warnings.

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

Warning
Location
SKILL.md:23
Finding

Public Static Salts and MD5 Used for Request Signing

Content
View full analysis

Vulnerability Details

File Locations:

  • SKILL.md:23
  • pms_m_editpushmsgreadstatus.md:3,13
  • pms_m_pushmsginfo.md:3,13
  • pms_m_pushmsglist.md:3,13
  • pms_u_editpushmsgreadstatus.md:3,13
  • pms_u_pushmsginfo.md:3,13
  • pms_u_pushmsglist.md:3,13
  • pms_uploadcustominfo.md:3,13

Vulnerability Type: Weak request-signing mechanism and exposed static signing material
Risk Level: Medium

Vulnerable Code Snippets

SKILL.md:23:

markdown
- **서명**: `sign = MD5(<params> + <salt>).toUpperCase()` — 엔드포인트별 salt는 각 문서 참조

pms_m_editpushmsgreadstatus.md:3,13:

yaml
description: "판매자 측에서 메시지를 읽음으로 표시. sign salt: aseftyl$ssk2"
markdown
판매자 측에서 메시지를 읽음으로 표시. sign salt: aseftyl$ssk2

pms_m_pushmsginfo.md:3,13 and pms_u_pushmsginfo.md:3,13:

yaml
description: "판매자 측 홈 공지 요약 조회. sign salt: asdjl$ssk2"
markdown
판매자 측 홈 공지 요약 조회. sign salt: asdjl$ssk2
yaml
description: "사용자 측 홈 공지 요약 조회. sign salt: asdjl$ssk2"
markdown
사용자 측 홈 공지 요약 조회. sign salt: asdjl$ssk2

pms_m_pushmsglist.md:3,13 and pms_u_pushmsglist.md:3,13:

yaml
description: "판매자 측 공지 목록 조회(페이지네이션). sign salt: sadasaa4r"
markdown
판매자 측 공지 목록 조회(페이지네이션). sign salt: sadasaa4r
yaml
description: "사용자 측 공지 목록 조회(페이지네이션). sign salt: sadasaa4r"
markdown
사용자 측 공지 목록 조회(페이지네이션). sign salt: sadasaa4r

pms_u_editpushmsgreadstatus.md:3,13:

yaml
description: "사용자 측에서 메시지를 읽음으로 표시. sign salt: aedrttyl$ssk2"
markdown
사용자 측에서 메시지를 읽음으로 표시. sign salt: aedrttyl$ssk2

pms_uploadcustominfo.md:3,13:

yaml
description: "APP에서 커스텀 정보 업로드. sign salt: asefimrvwl$ssk2"
markdown
APP에서 커스텀 정보 업로드. sign salt: asefimrvwl$ssk2

Technical Analysis

The documented signing algorithm calculates an unkeyed MD5 digest from request parameters and a static endpoint- ...[truncated 2205 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace the MD5 construction with HMAC-SHA-256 or a modern authenticated request-signing standard.
  2. Use a cryptographically random signing key that is issued per client or installation and is never embedded in Skill documentation, source control, logs, or client-visible metadata.
  3. Define an unambiguous canonical request containing the HTTP method, normalized path, canonical query parameters, body digest, timestamp, and nonce.
  4. Validate timestamps within a short server-side acceptance window and reject reused nonces to prevent replay attacks.
  5. Continue enforcing the access token independently on every non-public endpoint; never treat possession of a request signature alone as authorization.
  6. Apply least-privilege scopes so read-status updates and custom-information uploads require narrowly scoped write permissions.
  7. Rotate existing signing material after migration and invalidate legacy MD5 signatures.
  8. Add negative security tests covering altered parameters, replayed requests, expired timestamps, duplicate nonces, missing tokens, and tokens with insufficient scope.
  9. Avoid publishing cryptographic secrets in endpoint descriptions or metadata. Documentation should describe the protocol without containing operational key material.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (14)

Tp2

High
Category
MCP Tool Poisoning
Confidence
85% confidence
Finding

Mixing characters from multiple Unicode scripts in a single identifier is a common technique to create visually ambiguous tool names.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The documentation explicitly exposes a signing salt/secret (aseftyl$ssk2) in frontmatter and body text, which is sensitive material that should not be disclosed in a skill spec. If this value is used to generate or validate request signatures, an attacker or unauthorized integrator could forge trusted requests or reverse-engineer authentication flows beyond the intended push-message read-status operation.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill explicitly exposes a signing salt (asdjl$ssk2) in both front matter and body text. Publishing secret material in skill documentation can enable request forgery, signature bypass, or compromise of any authentication scheme that relies on the salt remaining confidential, and the repeated exposure increases the likelihood of misuse.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill metadata and description explicitly disclose a signing secret ('sign salt: sadasaa4r') in plaintext. Embedding secrets in documentation or skill definitions can enable unauthorized request signing, facilitate API abuse, and lead to further compromise if the same secret is reused elsewhere. The fact that this skill also documents the auth mechanism and endpoint makes the disclosure more actionable, not less.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The documentation embeds a signing secret (sign salt: asdjl$ssk2) directly in the skill metadata and body, which is a credential/secret exposure. Even though the endpoint is described as read-only, exposed signing material can enable unauthorized request signing, abuse of related APIs, or serve as a foothold if the same salt is reused elsewhere; the repeated inclusion in the file increases confidence this is an actual leak rather than placeholder text.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill advertises state-changing endpoints such as read-status updates and custom info uploads, but it does not warn users that invoking them will modify server-side data. In agent workflows, this can lead to unintended actions being triggered without clear user awareness or confirmation, especially when a tool is used autonomously or semi-autonomously.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The common header requirements specify lang=zh-cn as a fixed value. This is a natural-language/locale policy issue because it forces a specific language setting without user opt-in or explanation that the API is region- or locale-specific.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The usage instructions direct users to place an access token in the authcode header but provide no guidance on secure handling, storage, redaction, or logging avoidance. In agent environments, this increases the risk of credential exposure through prompts, traces, logs, screenshots, or copied example commands.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown file describes a write-permission POST operation that changes message state to 'read', which can affect message integrity or audit semantics. The description presents the action as a simple reference but does not include any warning that invoking it will alter data or should only be used intentionally.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The file content does not align with the declared APM push-message service context, instead describing a seller home-notice summary endpoint. In security-sensitive API skill definitions, this kind of semantic mismatch can mislead operators, reviewers, or downstream agents into calling the wrong endpoint or granting trust to functionality that was not actually vetted.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The documentation explicitly exposes a hardcoded signing salt ('sadasaa4r') even though this skill only describes a user-facing read/list API. Publishing secret signing material in docs weakens any request-signing scheme that depends on that salt and can enable unauthorized request forgery or easier reverse engineering of adjacent protected endpoints. In this context, the issue is more dangerous because the skill also documents a private, non-public API and authentication conventions, making the leaked salt operationally useful rather than purely illustrative.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill documents a write-capable POST upload endpoint that transmits custom information but provides no warning that invoking it will send or modify data. This creates a real safety issue for agent-driven use because users or downstream systems may trigger state-changing uploads without clear consent, and the embedded signing salt further increases sensitivity by encouraging direct request construction.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

This markdown file describes a POST endpoint with write permission that changes message read status, which affects user data/state. The description states what the endpoint does but does not include any caution or user-facing warning about modifying message state.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

The file's user-facing name, description, and category are presented only in Korean, with no indication that users may choose another language or locale. Under the policy rule, a language constraint should either be optional or explicitly justified as region-specific.

Content

No source excerpt is available for this finding.