Back to skill

Security audit

Skill Amazon Ads Optimizer

Security checks for vulnerabilities and agentic risk

Overview

This Amazon Ads skill mostly does what it says, but it asks users to store durable ad-account credentials in a local plaintext file without adequate safety guidance or containment.

Review before installing. Use a dedicated least-privilege Amazon Ads app/token, keep the credentials file outside repositories and shared workspaces, set restrictive file permissions, add it to .gitignore, and rotate the token if it may have been exposed. Treat campaign exports as sensitive business data.

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
scripts/ads.js:12
Finding
Long-Lived Amazon Credentials Stored in an Unprotected Plaintext File## Vulnerability Details **File Location**: `scripts/ads.js:12-16`; related setup instructions in `SKILL.md:15-24` and `README.md:22-31` **Vulnerability Type**: Plaintext sensitive credential storage **Risk Level**: Medium ### Vulnerable Code ```js const CREDS_PATH = process.env.AMAZON_ADS_PATH || './amazon-ads-api.json'; const ENDPOINTS = { na: 'advertising-api.amazon.com', eu: 'advertising-api-eu.amazon.com', fe: 'advertising-api-fe.amazon.com' }; function getCreds() { return JSON.parse(fs.readFileSync(CREDS_PATH, 'utf8')); } ``` The documented credential file contains the following long-lived secrets: ```json { "lwaClientId": "amzn1.application-oa2-client.YOUR_CLIENT_ID", "lwaClientSecret": "YOUR_CLIENT_SECRET", "refreshToken": "Atzr|YOUR_REFRESH_TOKEN", "profileId": "YOUR_ADS_PROFILE_ID", "region": "eu" } ``` ### Technical Analysis The Skill requires users to save an Amazon Login with Amazon client secret and refresh token in a plaintext JSON file. By default, that file is expected in the current project directory as `./amazon-ads-api.json`. Neither the implementation nor the documentation enforces or recommends restrictive file permissions, storage outside the repository, use of a secret manager, or exclusion from version control. The project also does not include a `.gitignore` rule protecting the documented filename. A refresh token is a long-lived credential that can be exchanged for access tokens. Consequently, exposure of this file is materially more serious than exposure of a short-lived access token. The code does not itself transmit these credentials to an unauthorized party: it sends them to Amazon's official HTTPS OAuth endpoint, which is necessary for the declared functionality. The weakness is the local storage and handling model. ### Attack Path 1. A user follows the setup instructions and creates `amazon-ads-api.json` in the project or working directory. 2. The file inhe ...[truncated 1346 chars]
Remediation
## Remediation Suggestions 1. Prefer an operating-system credential store, managed secret service, or protected environment variables over a project-local JSON file. 2. If file-based credentials must remain supported: - Require the file to be outside the repository by default. - Reject files whose permissions allow group or world access on supported platforms. - Recommend and enforce mode `0600` where practical. - Resolve the path explicitly and reject unexpected file types such as symbolic links if the execution environment is not trusted. 3. Add `amazon-ads-api.json` and common credential filename variants to `.gitignore`. 4. Update `README.md` and `SKILL.md` with explicit warnings against committing, logging, sharing, or backing up the credential file insecurely. 5. Document credential revocation and rotation procedures. 6. Configure the Amazon application and refresh token with only the minimum API permissions required to list profiles and campaigns. 7. Avoid including token endpoint response bodies in errors where they might contain sensitive diagnostic data; return a sanitized error code and message instead.

T09 · Insecure Skill Coding Practices

Note
Location
scripts/ads.js:125
Finding
Module Import Unconditionally Executes CLI Control Flow## Vulnerability Details **File Location**: `scripts/ads.js:125-126` **Vulnerability Type**: Import-time side effects and unintended network execution **Risk Level**: Low ### Vulnerable Code ```js main().catch(e => { console.error(e.message || e); process.exit(1); }); module.exports = { getProfiles, getCampaigns, getAccessToken }; ``` Relevant command processing invoked by `main()` includes: ```js async function main() { const args = parseArgs(); if (!args.command) { console.log('Usage:'); console.log(' node ads.js --profiles'); console.log(' node ads.js --campaigns [--out file.json]'); console.log(' node ads.js --summary'); process.exit(0); } ``` ### Technical Analysis The file exports reusable functions through `module.exports`, but it also calls `main()` unconditionally. Node.js executes top-level statements when a module is loaded with `require()`. Therefore, importing this module triggers CLI argument parsing and may terminate the entire host process or initiate credential reads and Amazon API requests. This violates the expected separation between reusable library behavior and executable CLI behavior. The issue is especially relevant in an agent host, test runner, or larger automation process where importing the advertised functions should not execute unrelated control flow. The reachable network destinations remain hard-coded official Amazon HTTPS endpoints. No attacker-selected destination or remote code execution was identified. ### Attack Path 1. Another component imports `scripts/ads.js` to call one of the exported functions. 2. Node.js evaluates the module and immediately invokes `main()`. 3. If no recognized command is present, `main()` invokes `process.exit(0)`, terminating the importing process. 4. Alternatively, if the host process arguments contain `--profiles`, `--campaigns`, or `--summary`, the imported module reads the configured credential file. ...[truncated 1034 chars]
Remediation
## Remediation Suggestions Guard CLI execution so it occurs only when the script is launched directly: ```js if (require.main === module) { main().catch(e => { console.error(e.message || e); process.exitCode = 1; }); } module.exports = { getProfiles, getCampaigns, getAccessToken }; ``` In addition: 1. Avoid calling `process.exit()` inside reusable control flow; set `process.exitCode` or return an explicit status instead. 2. Separate CLI parsing into a dedicated entry-point file and keep API functions in a side-effect-free module. 3. Add tests confirming that importing the module performs no filesystem access, network requests, console output, file writes, or process termination. 4. Validate that `--out` has a following value and document that it writes potentially sensitive campaign information.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (12)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The description is partially accurate: the script does list profiles and show campaign/budget summaries for Amazon Ads Sponsored Products. However, it overstates the functionality in material ways. There is no campaign management capability implemented, only a POST to /sp/campaigns/list and a profiles fetch. The reported 'performance' is not actual ad performance data; the script only summarizes campaign states and daily budgets from campaign metadata. Additionally, operation is tied to configured credentials and a profileId in a local JSON file, so 'works with any advertiser account' is broader than what the code demonstrates. The file output capability is a supporting detail, not a mismatch by itself.

Ae1

High
Category
analysis-evasion
Content
node scripts/ads.js --profiles
Confidence
100% confidence
Finding
Referenced artifact was not completely inspected

Ae1

High
Category
analysis-evasion
Content
node scripts/ads.js --profiles
Confidence
100% confidence
Finding
Referenced artifact was not completely inspected

Ae1

High
Category
analysis-evasion
Content
node scripts/ads.js --profiles
Confidence
100% confidence
Finding
Referenced artifact was not completely inspected

Ae1

High
Category
analysis-evasion
Content
node scripts/ads.js --profiles
Confidence
100% confidence
Finding
Referenced artifact was not completely inspected

Ae1

High
Category
analysis-evasion
Content
node scripts/ads.js --profiles
Confidence
100% confidence
Finding
Referenced artifact was not completely inspected

Ae1

High
Category
analysis-evasion
Content
node scripts/ads.js --profiles
Confidence
100% confidence
Finding
Referenced artifact was not completely inspected

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The README instructs users to place long-lived Amazon Ads credentials, including client secret and refresh token, in a local JSON file without any warning about secure storage, exclusion from source control, or file permission hardening. In an agent-skill context, this increases the chance of accidental credential exposure through commits, logs, workspace sharing, or multi-agent access, which could enable unauthorized access to Amazon Ads accounts.

Lp3

Medium
Category
MCP Least Privilege
Confidence
87% confidence
Finding
The skill declares no explicit tool scope or permission boundaries even though its documented behavior requires environment access and outbound network calls. In an agent setting, missing scope declarations can cause overbroad execution privileges, making it harder to constrain or audit access to secrets and external APIs.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The setup instructs users to store high-value credentials, including client secret and refresh token, in a local JSON file and reference it via an environment variable path without warning about secret-handling risks. This increases the chance of accidental exposure through source control, weak filesystem permissions, backups, logs, or misconfigured shared environments.

External Transmission

Medium
Category
Data Exfiltration
Content
async function getAccessToken() {
  const creds = getCreds();
  const res = await fetch('https://api.amazon.com/auth/o2/token', {
    method: 'POST',
    headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
    body: new URLSearchParams({
Confidence
70% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

External Transmission

Medium
Category
Data Exfiltration
Content
async function getAccessToken() {
  const creds = getCreds();
  const res = await fetch('https://api.amazon.com/auth/o2/token', {
    method: 'POST',
    headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
    body: new URLSearchParams({
Confidence
60% confidence
Finding
Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Static analysis

Detected: suspicious.env_credential_access, suspicious.exposed_secret_literal

Environment variable access combined with network send.

Critical
Code
suspicious.env_credential_access
Location
scripts/ads.js:12

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
scripts/ads.js:32