Back to skill

Security audit

Jarvis Money Maker

Security checks for vulnerabilities and agentic risk

Overview

This skill is only guidance text, but it encourages automated account activity with local credential files and recurring/sub-agent work without enough user control or safety boundaries.

Review before installing. Do not store real credentials in project JSON files unless they are encrypted, excluded from source control, and narrowly scoped. Require explicit confirmation before any agent posts, upvotes, submits tasks, registers accounts, or runs recurring automation, and define a clear stop condition.

Vulnerability Patterns
  • System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
  • 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

Warning
Location
SKILL.md:136
Finding

Plaintext Credential Storage in JSON Files

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, line 136
Vulnerability Type: Plaintext sensitive-data storage
Risk Level: Medium

Vulnerable Code Snippet:

markdown
- Store credentials in `credentials/*.json`

Technical Analysis

The Skill explicitly recommends storing credentials in JSON files but does not require encryption, restrictive filesystem permissions, repository exclusion, credential scoping, or integration with a secret manager. JSON files provide no confidentiality protection by themselves.

Credentials stored in this manner may be exposed through source-control commits, shared workspaces, backups, diagnostic archives, logs, or access by other local users and processes. Because the Skill directs users to register and operate accounts on several external platforms, exposed credentials could grant access to those accounts and their associated API capabilities.

Attack Path

  1. An agent follows the Skill and writes third-party API credentials to a file under credentials/.
  2. The file remains unencrypted and is not protected by an explicitly required restrictive permission policy.
  3. The file is accidentally committed to source control, included in a backup or diagnostic archive, shared with another agent, or read by another local process.
  4. An unauthorized party obtains the stored token or credential.
  5. The party reuses it against the relevant external service to act with the permissions assigned to that credential.

Impact Assessment

Exploitation could permit unauthorized access to affected third-party accounts. The precise scope depends on each exposed credential, but it may include reading account information, creating posts, submitting tasks, publishing content, voting, or performing other API operations available to the account. The issue does not, by itself, demonstrate operating-system privilege escalation or compromise beyond the permissions granted to the exposed c ...[truncated 11 chars]

Remediation
View remediation

Remediation Suggestions

  • Replace plaintext JSON credential storage with an operating-system, cloud, or dedicated application secret manager.
  • If file-based storage is unavoidable, encrypt credentials at rest and keep decryption keys outside the project directory.
  • Require restrictive filesystem permissions so only the intended user or service account can read the credential file.
  • Add credential paths to .gitignore and enforce secret scanning in local hooks and continuous-integration pipelines.
  • Use narrowly scoped, short-lived API tokens rather than reusable account passwords or broad tokens.
  • Prohibit printing credentials in logs, prompts, generated reports, or agent transcripts.
  • Document credential rotation and revocation procedures and rotate any secret suspected of exposure.

T06 · System Persistence

Warning
Location
SKILL.md:132
Finding

Unbounded Recurring and Delegated Agent Automation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 132-136
Vulnerability Type: Persistent and insufficiently constrained automation
Risk Level: Medium

Vulnerable Code Snippet:

markdown
## Automation Tips

- Use sub-agents with `sessions_spawn` for parallel work
- Set up HEARTBEAT.md for periodic checks
- Store credentials in `credentials/*.json`

Technical Analysis

The Skill recommends spawning sub-agents and establishing periodic checks without defining authorization gates, permitted operations, credential boundaries, execution budgets, expiration conditions, or a shutdown procedure. In the broader context of the Skill, the delegated work can involve account registration, task submission, content publication, posting, and voting on third-party platforms.

Periodic automation can continue beyond the immediate interactive request if the surrounding agent framework treats HEARTBEAT.md as persistent scheduling or recurring-session configuration. Parallel sub-agents also increase the number of execution contexts that may access credentials or perform external actions. The text does not explicitly install a backdoor or guarantee that recurring execution will occur; the risk arises when these recommendations are implemented by a framework that supports persistent heartbeat tasks.

Attack Path

  1. A user activates the Skill to automate monetization-related activity.
  2. The primary agent follows the recommendation to create parallel sub-agent sessions.
  3. It configures periodic checks through HEARTBEAT.md and gives the automation access to third-party account credentials.
  4. No explicit action allowlist, per-action approval requirement, execution deadline, or resource budget limits subsequent activity.
  5. Sub-agents or recurring checks continue making account-related requests, submissions, or posts after the initiating interaction or after the user's intent changes.
  6. The automation co ...[truncated 677 chars]
Remediation
View remediation

Remediation Suggestions

  • Require explicit, informed user approval before spawning sub-agents or configuring recurring checks.
  • Define a strict allowlist of permitted platforms, endpoints, account operations, and data accessible to each sub-agent.
  • Require separate confirmation for externally visible or account-changing actions such as registration, publication, posting, voting, and submissions.
  • Assign narrowly scoped credentials to each worker rather than sharing broad account credentials.
  • Add expiration times, maximum run counts, API-request budgets, concurrency limits, and platform-specific rate limits.
  • Make periodic behavior read-only by default and prevent recurring jobs from performing account-changing actions without renewed approval.
  • Record attributable audit logs for every spawned session and external action without recording secrets.
  • Provide a documented shutdown mechanism that terminates child sessions, removes recurring configuration, and revokes temporary credentials.
  • Clearly disclose whether HEARTBEAT.md survives the current session before creating or modifying it.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly recommends storing credentials in credentials/*.json, which encourages local plaintext or weakly protected secret storage without any guidance on encryption, access controls, or secret-management tooling. In a monetization/automation skill that integrates multiple external platforms, those credentials likely grant account access and API capabilities, so disclosure could lead to account takeover, fraud, or abuse across several services.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The package description contains Chinese text indicating the skill is presented in a specific language, but the file provides no opt-in, language choice, or justification that it is intended only for a Chinese-speaking or region-specific audience. This can violate language/locale policy when users are not given a choice.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The description includes Chinese-language trigger text and branding while the rest of the document is primarily in English. This creates a language/locale constraint or expectation without explicitly offering the user a language preference or documenting the locale choice.

Content

No source excerpt is available for this finding.