Back to skill

Security audit

Feima Lab Content Manager

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed feima-lab content manager, but its remote publishing and image workflows create review-worthy risks for accidental backend changes and unintended file or network exposure.

Install only if you trust the publisher and intend to let the agent manage feima-lab backend content with an internal API key. Keep the key in the shell only, review every remote save/publish/unpublish/upload action before it runs, and avoid processing untrusted MDX or image URLs until path containment and outbound-fetch restrictions are added.

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

Error
Location
scripts/api/save-article.mjs:206
Finding

Path Traversal Allows Upload of Arbitrary Readable Local Files

Content
View full analysis
/images`. The resolved path is passed to `apiPostMultipart()`, which reads the entire file and uploads it to the configured feima-lab endpoint. Symbolic links placed inside the images directory can create an equivalent escape because the code does not compare canonical paths obtained through `realpath()`. Although uploading article images is necessary for the declared publishing functionality, reading and uploading files outside the article image directory exceeds the minimum filesystem privileges requir ...[truncated 1426 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/image-localize.mjs:25
Finding

Unrestricted Image Fetching Enables Server-Side Request Forgery

Content
View full analysis
controller.abort(), 15000); const res = await fetch(url, { signal: controller.signal }); clearTimeout(timer); if (!res.ok) throw new Error(`HTTP ${res.status}`); const ext = extFromContentType(res.headers.get('content-type')); const actualPath = destPath.endsWith('.bin') ? destPath.replace(/\.bin$/, ext) : destPath; const buf = Buffer.from(await res.arrayBuffer()); await writeFile(actualPath, buf); return actualPath; } catch (e) { lastErr = e; } } throw lastErr; } async function resolveImage(ref, mdxDir, imagesDir) { if (ref.startsWith('./images/') || ref.startsWith('images/')) return { skipped: true, newRef: ref }; if (/^https?:\/\//.test(ref)) { const urlObj = new URL(ref); const bn = basename(urlObj.pathname) || 'img'; const hash = sha1short(ref); const tempName = `${hash}-${bn.replace(/\.[^.]+$/, '')}.bin`; const destPath = join(imagesDir, tempName); const actualPath = await downloadWithRetry(ref, destPath); return { skipped: false, newRef: `./images/${basename(actualPath)}` }; } ``` ### Technical Analysis The image localization feature passes URLs extracted from article content directly to `fetch()`. The implementation does not restrict destination hosts or resolved IP addresses and therefore permits requests to: - Loopback addresses - Private network ranges - Link-local addresses - Internal DNS names - Local development services - Cloud metadata or orchestration endpoints - Other network locations accessible from the host ...[truncated 2132 chars]
Remediation
View remediation
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (48)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 38)May include surrounding context.

md
- `scripts/render.mjs` 是预打包的自包含文件(包含 unified/remark/mdast/hast 等所有 MDX 解析依赖)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 119)May include surrounding context.

md
- `scripts/render.mjs` 是预打包的自包含文件(包含 unified/remark/mdast/hast 等所有 MDX 解析依赖)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 39)May include surrounding context.

md
- `scripts/new-post.mjs` / `scripts/image-localize.mjs` / `scripts/api/*` 只用 Node 18+ 内置 API(fetch / FormData / Blob 等)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 116)May include surrounding context.

md
- `scripts/new-post.mjs` / `scripts/image-localize.mjs` / `scripts/api/*` 只用 Node 18+ 内置 API(fetch / FormData / Blob 等)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 127)May include surrounding context.

md
node scripts/api/list-tags.mjs [--format json|table]

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 128)May include surrounding context.

md
node scripts/api/get-article.mjs --slug <slug>

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 129)May include surrounding context.

md
node scripts/api/get-article.mjs --slug <slug>

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 133)May include surrounding context.

md
node scripts/api/save-article.mjs --post-dir <posts/slug> [--dry-run]

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 135)May include surrounding context.

md
node scripts/api/publish-article.mjs --post-dir <posts/slug> | --id <n> | --slug <slug>

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 136)May include surrounding context.

md
node scripts/api/unpublish-article.mjs --post-dir <posts/slug> | --id <n> | --slug <slug>

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The workflow explicitly requires an INTERNAL API key for backend operations, which indicates privileged access beyond normal end-user scope. Embedding privileged operational guidance inside a content-authoring skill is dangerous because it normalizes use of elevated credentials for routine tasks and could lead to unauthorized publication, modification, or metadata changes if the skill is triggered inappropriately.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README describes powerful remote capabilities including article CRUD, publishing, tag creation, and file upload, but it does not prominently warn that these operations modify remote systems and may create or overwrite content. In an agent setting, omission of this warning makes unsafe default use more likely, especially when paired with conversational triggers that may invoke these actions implicitly.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The README advertises very broad natural-language triggers for remote publishing such as 'push to remote' and 'publish this to feima-lab', which can plausibly overlap with ordinary user conversation and cause the agent to invoke state-changing actions without sufficiently explicit intent. In a skill that supports publish, unpublish, CRUD, tag creation, and uploads, ambiguous activation materially increases the risk of accidental remote modification or disclosure.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
94% confidence
Finding

The skill exposes capabilities that can read environment variables and perform networked API operations, including saving, publishing, unpublishing, and uploading content, but it does not declare an explicit tool scope or allowed-tools boundary. That makes the operational blast radius ambiguous and increases the chance the agent can invoke powerful side-effecting actions without clear policy constraints or user visibility.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger list contains broad phrases like blog writing, publishing, saving articles, listing content, and style conversion that are likely to match ordinary user requests beyond the intended narrow context. Overbroad activation can cause this skill to be selected unexpectedly, increasing the chance of unintentional file manipulation or remote API actions in situations where the user did not mean to invoke feima-lab publishing workflows.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill advertises remote upload, save, publish, unpublish, and OSS image upload capabilities, but the description does not prominently warn that these actions can create external side effects on remote systems. In practice, a user may invoke what appears to be formatting or content assistance and unknowingly trigger state-changing operations against production-like content infrastructure.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documentation explicitly tells the agent to show the user the complete unexpected error JSON, and it notes that the message may contain a stack trace. Stack traces and raw exception payloads can disclose filesystem paths, internal code structure, request details, environment-derived values, or other sensitive debugging context that should not be broadly exposed.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The entire operational guidance, user-facing progress messages, and prescribed interaction patterns are specified only in Chinese, including exact status text the agent should emit. There is no indication that the user can choose another language or that the Chinese-only constraint is limited to a justified region-specific context.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The documented behavior expands the skill from local content authoring into remote backend operations: uploading files, saving articles, managing tags/categories, and publishing state changes. That is a material scope increase over the stated writing/conversion purpose, which raises confused-deputy risk: a user invoking a 'content manager' may unknowingly authorize privileged actions against a backend.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The workflow includes publication control actions such as publish and unpublish, which directly change production content visibility. In the context of a writing/formatting skill, this is dangerous because a user may intend only document preparation while the skill can alter live backend state, enabling accidental or unauthorized takedowns and releases.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

该 markdown 文件全文以中文规定组件选择规则,未见提供语言/地区选择、用户 opt-in,或说明该限制仅适用于特定中文文档场景。根据规则,强制特定语言而无明确选择机制,属于自然语言层面的语言/locale 策略问题。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The entire skill documentation is written in Chinese, including the title, purpose, rules, and examples, with no indication that the skill is region-specific or that other languages are supported. Under the policy, forcing a specific language without user opt-in or justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The entire skill reference is written in Chinese and does not indicate that users may choose another language or that the component is intended only for a Chinese-language environment. Under the policy, forcing a specific language without user opt-in is a natural-language locale violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The entire skill documentation is written only in Chinese, including headings, usage notes, and rules, with no indication that language choice is optional or that the component is intentionally limited to a Chinese-only context. This can violate language/locale policy when users are not given an opt-in or alternative locale.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

This markdown file presents all user-facing instructions in Chinese, and there is no indication that the skill is region-specific or that users may choose another language. Under the policy, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.