Back to skill

Security audit

Online Resume Creator

Security checks for vulnerabilities and agentic risk

Overview

The skill mostly does what it claims, but crafted resume data can inject JavaScript into the generated public resume page.

Install only if you trust the resume JSON/content you feed into it, use a virtual environment for Pillow/qrcode, and host generated pages on an isolated static origin. Do not process third-party resume JSON until the script escapes data safely for inline scripts and validates input fields.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T09 · Insecure Skill Coding Practices

Error
Location
scripts/make_resume.py:72
Finding

Stored JavaScript Injection Through Unsafely Embedded Resume Data

Content
View full analysis
{section_js} {rules_js} const chat = document.getElementById('chat'); ``` ### Technical Analysis The `sections` and `rules` values can originate from a user-supplied JSON file loaded through the `--data` option. Although `json.dumps()` correctly escapes data for a JavaScript string literal, ordinary JSON encoding does not make the result safe for direct placement inside an HTML `` terminates a script element even when it occurs inside what JavaScript would otherwise treat as a quoted string. Consequently, a crafted resume value can close the existing script element and introduce a new executable element. For example, an attacker-controlled section value could contain: ```html ``` When the generated resume is opened, the browser parses the injected closing tag before JavaScript string-literal semantics apply. The attack therefore becomes stored JavaScript injection in the generated HTML file. Using `d.textContent` when displaying answers does not mitigate this issue because execution occurs while the browser initially parses the generated script block, before the answer is rendered. ### Attack Path 1. An attacker creates or modifies a resume JSON file containing a malicious `sections` value or rule value with a `` sequence and an injected script element. ...[truncated 1437 chars]
Remediation
View remediation
", "\\u003e") .replace("\u2028", "\\u2028") .replace("\u2029", "\\u2029") ) section_js = "const SECTION = " + json_for_script(sections) + ";" rules_js = "const RULES = " + json_for_script( [{"k": k, "w": w} for k, w in rules] ) + ";" ``` Escaping `<` prevents attacker-controlled data from producing a literal `` sequence in the generated HTML. Additional hardening should include: 1. Validate the structure and types of all fields loaded from `--data`. 2. Reject unexpected nested objects or non-string keywords and section values. 3. Add automated tests using payloads containing ``, HTML tags, quotes, ampersands, and U+2028/U+2029. 4. Consider placing serialized data in a dedicated JSON data element with safe escaping, then parsing its text content. 5. Deploy generated pages on an isolated origin rather than an origin shared with authenticated or sensitive applications. 6. Add a restrictive Content Security Policy as defense in depth, preferably using hashes or nonces rather than allowing unrestricted inline scripts. ]]>

T08 · Insecure Dependencies

Note
Location
references/quickstart.md:8
Finding

Unpinned Third-Party Python Dependencies

Content
View full analysis
Remediation
View remediation
qrcode== ``` 2. Generate and publish cryptographic hashes for all direct and transitive packages, then require them during installation: ```bash python -m pip install --require-hashes -r requirements.txt ``` 3. Use a lock-file workflow that records transitive dependencies and review lock-file changes before release. 4. Document the intended trusted package index and avoid untrusted `--extra-index-url` configurations. 5. Install dependencies inside a dedicated virtual environment as a non-administrative user. 6. Add automated dependency vulnerability and provenance scanning to the release process. 7. Update `manifest.yaml` so its dependency declarations correspond to the reviewed, pinned dependency set. ]]>
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (6)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The code does match the core resume-page behavior: it builds a single-file HTML resume with an embedded conversational assistant implemented via client-side keyword matching, and it can load resume data from JSON. However, the declared description also claims creation of WeChat sharing assets (cover image and QR-code share card) and deployability as a shareable link. None of those capabilities appear in the supplied code. The script only writes a local HTML file; it does not create images, QR codes, or perform deployment/link publishing. Therefore the description overstates the implemented functionality in material ways.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill describes workflows that read user-provided resume content and generate files such as HTML and images, but it does not declare any explicit tool scope or allowed-tools boundary. Without scoped permissions, an agent platform may grant broader filesystem access than users expect, increasing the risk of unintended reads or writes when handling sensitive personal resume data.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description says to use the skill when the user wants 'a personal homepage/portfolio' or for 'personal branding,' which are broad intents that can match many ordinary web-page requests beyond this skill's narrow resume use case. It does not provide exclusion conditions or negative examples to clarify when a general homepage request should not trigger this skill.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The documentation states that visitors ask questions '用中文提问' (ask in Chinese), which establishes a Chinese-only interaction expectation. There is no indication that users can choose another language or opt in to this locale restriction, making it a natural-language policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The description states that visitors ask questions in Chinese, which imposes a specific language expectation in the skill's natural-language behavior. There is no opt-in, alternative language option, or explanation that this is intentionally limited to a China-specific or Chinese-only use case.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The generated HTML sets lang="zh-CN", and the surrounding UI and assistant prompts are all fixed in Chinese. Under the policy, forcing a specific language without user opt-in is a natural-language locale violation unless clearly documented as a justified region-specific tool.

Content

No source excerpt is available for this finding.