Back to skill

Security audit

offer-tracker

Security checks for vulnerabilities and agentic risk

Overview

This skill is mostly a local job-application tracker, but it has unsafe data-handling and overwrite behavior that users should review before installing.

Install only if you are comfortable reviewing or fixing the HTML template first. Do not import JSON backups from untrusted sources, avoid opening exported CSVs in spreadsheets without checking formula-like cells, and ask for confirmation before the agent creates or overwrites Desktop tracker and backup files.

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 (3)

T09 · Insecure Skill Coding Practices

Error
Location
assets/application-tracker.html:305
Finding

Persistent DOM-Based Cross-Site Scripting Through Imported Record IDs

Content
View full analysis
` ${escapeHtml(formatDate(item.date))} ${escapeHtml(item.company)} ${escapeHtml(item.position || '-')} ${escapeHtml(item.status)} ${escapeHtml(item.next || '-')} ${escapeHtml(item.note || '-')}
编辑删除
`).join(''); ``` The corresponding import validation is: ```javascript let id = item.id; if (id == null || id === '') { do { id = makeId(); } while (seenIds.has(id)); } else if (typeof id !== 'string' || !id.trim() || id !== id.trim()) fail('ID 格式无效'); if (seenIds.has(id)) fail('ID 重复'); seenIds.add(id); ``` ### Technical Analysis The application renders records by assigning an HTML string to `body.innerHTML`. Most displayed record properties are passed through `escapeHtml`, but `item.id` is inserted directly into the quoted `data-edit` and `data-delete` attributes. Imported IDs are only checked for type, surrounding whitespace, emptiness, and duplication. Quotes, angle brackets, and event-handler syntax are accepted. An attacker can therefore terminate the `data-edit` attribute and inject another attribute or element. For example, an imported ID similar to the following can create an executable event handler: ```text " onmouseover="alert(document.domain) ``` Because imported records are saved in `localStorage`, the injected markup is persistent and is rendered again wheneve ...[truncated 1399 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:52
Finding

Stored Script Injection Through Unsafe Embedding of User Records in Inline JavaScript

Content
View full analysis
` element. HTML parsing recognizes a literal `` sequence even when it appears inside what JavaScript would otherwise treat as a quoted string. A malicious value can therefore terminate the existing script element and inject new HTML or script content. Unsafe manual object construction can also permit direct JavaScript string termination if quotes, backslashes, line separators, or other control characters are not encoded correctly. ### Attack Path 1. A user or attacker supplies a job-application field containing script-context termination content, such as a crafted note containing a closing script tag followed by attacker-controlled markup. 2. The agent follows the Skill workflow and appends the record to `initialRecords` in the HTML file. 3. The value is inserted without mandatory script-context-safe serialization and escaping. 4. The generated tracker is opened in a browser. 5. ...[truncated 860 chars]
Remediation
View remediation
[] ``` 4. Serialize the complete data structure with a trusted JSON serializer and then encode characters that are significant in HTML script contexts. At minimum, encode `<`, `>`, `&`, U+2028, and U+2029: ```javascript const safeJson = JSON.stringify(records) .replace(//g, '\\u003e') .replace(/&/g, '\\u0026') .replace(/\u2028/g, '\\u2028') .replace(/\u2029/g, '\\u2029'); ``` 5. Parse the non-executable data with `JSON.parse(document.getElementById('initial-records').textContent)`. 6. Update the Skill instructions to prohibit manual string concatenation and require one well-defined serialization routine. 7. Add tests using quotes, backslashes, newlines, closing script tags, HTML tags, and Unicode line separators in every free-form field. 8. Apply a restrictive Content Security Policy as defense in depth. ]]>

T09 · Insecure Skill Coding Practices

Warning
Location
assets/application-tracker.html:345
Finding

Spreadsheet Formula Injection in CSV Export

Content
View full analysis
[item.date, item.time, item.company, item.position, item.status, item.next, item.note]); const csv = [header, ...rows].map((row) => row.map((cell) => `"${String(cell ?? '').replace(/"/g, '""')}"`).join(',')).join('\n'); download(`秋招投递进度-${today()}.csv`, '\ufeff' + csv, 'text/csv;charset=utf-8'); } ``` ### Technical Analysis The CSV exporter escapes double quotes and wraps every value in quotes. This protects the CSV structure but does not prevent spreadsheet applications from treating field values as formulas. A value whose first significant character is `=`, `+`, `-`, or `@` may be interpreted as a formula when the exported file is opened in Excel, LibreOffice Calc, or another compatible application. Such values can enter the tracker through manual input or imported JSON. CSV quoting is not a formula-neutralization mechanism. Spreadsheet software commonly evaluates quoted CSV cells after parsing them. ### Attack Path 1. An attacker creates or supplies a record containing a formula-like value in a field such as company, position, next step, or note. 2. The record is entered manually or imported from JSON. 3. The user selects the CSV export function. 4. The exporter wraps the value in double quotes but leaves the formula marker unchanged. 5. The user opens the exported CSV in spreadsheet software. 6. The spreadsheet interprets the cell as a formula. 7. Depending on the spreadsheet and its security settings, the formula can make external requests, display deceptive links, disclose spreadsheet data, or invoke dangerous legacy functionality. ### Impact Assessment The vulnerability crosses the trust boundary from th ...[truncated 543 chars]
Remediation
View remediation
row.map(safeCsvCell).join(',')) .join('\n'); ``` 3. Apply the protection to every exported field rather than only fields considered likely to contain free-form text. 4. Document that imported data is untrusted and that exported files should not contain active formulas. 5. Add regression tests for values beginning with formula markers, leading whitespace followed by a marker, tabs, carriage returns, and Unicode whitespace. 6. Where interoperability permits, consider exporting a format that explicitly stores values as text instead of relying on CSV type inference. ]]>
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (7)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 15)May include surrounding context.

md
- 进度表模板:`assets/application-tracker.html`(自带 `initialRecords` 数据数组,浏览器打开即渲染,数据存于 localStorage)。

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 25)May include surrounding context.

md
- 进度表模板:`assets/application-tracker.html`(自带 `initialRecords` 数据数组,浏览器打开即渲染,数据存于 localStorage)。

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest says to use the skill not only for explicit commands like /offer and /offer-tracker, but also whenever the user submits batch 'company-position-status' job records. That natural-language trigger scope is broad and lacks negative examples or tighter constraints, increasing the risk of unintended invocation during general discussion of applications.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The description states that this is a '中文秋招/实习求职进度管理技能' and the parsing rules are designed around Chinese input formats, but it does not present this as an opt-in language preference or provide alternatives. That can be a natural-language policy issue when a skill effectively forces a specific language or locale without explicit user choice.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill instructs creating and overwriting files on the user's Desktop, including a full JSON backup overwrite, without an explicit confirmation or warning step. This creates a real risk of unintended file creation, replacement, or data loss if the path already contains user-managed files or if the skill is triggered accidentally.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The document declares lang="zh-CN" and immediately presents the title and interface text only in Chinese. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation unless the locale limitation is explicitly justified or the user is offered a choice.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

This reference file mandates Chinese-language instructions, samples, and output text throughout, including the user-facing output template. Under the policy rule, forcing a specific language without opt-in is a natural-language locale violation unless the constraint is explicitly justified or users are offered a choice.

Content

No source excerpt is available for this finding.