Back to skill

Security audit

Ai Automation Agency Kit

Security checks for vulnerabilities and agentic risk

Overview

This is a prompt-only business kit, but one delivery prompt encourages handling client/admin credentials without safe transfer or revocation safeguards.

Review this skill carefully before installing. It appears to be a normal prompt pack, but users should not paste or send passwords, API keys, recovery codes, or admin credentials through generated emails, chat, or shared folders. Use named least-privilege accounts, password-manager or encrypted expiring secret-sharing workflows, credential rotation, MFA, access logs, and explicit agency-access revocation at handoff.

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 (1)

T09 · Insecure Skill Coding Practices

Warning
Location
prompts/prompt4-delivery-fulfillment-playbook.md:118
Finding
Insecure Transfer of Client Administrative Credentials## Vulnerability Details **File Location**: `prompts/prompt4-delivery-fulfillment-playbook.md`, lines 118–121 **Vulnerability Type**: Plaintext sensitive-data handling and insecure credential transfer **Risk Level**: Medium ### Vulnerable Code Snippet ```text a) Project close email (send day of final delivery): - Celebrate the win (specific results from the results report) - Provide everything: links, credentials, documentation folder - Set expectation for 30-day check-in ``` Related instructions reinforce handling and transfer of administrative credentials: ```text Min 40-55: access and logistics (accounts, credentials, shared drive) ``` ```text "Before moving to Launch, client approves: (1) all workflows run in test mode with no errors, (2) dashboard shows correct data for past 7 days, (3) client admin credentials are transferred and tested." ``` ### Technical Analysis The prompt instructs the generated project-close email to provide client credentials alongside links and documentation. It does not require a password manager, an expiring secret-sharing mechanism, encryption, a separate communication channel, credential rotation, multifactor authentication, or least-privilege service accounts. Email messages and project documentation are frequently retained, forwarded, synchronized to multiple devices, backed up, and accessible to mailbox administrators. Including administrative credentials in such communications can therefore create a long-lived plaintext secret. The related instructions also normalize transferring and testing client administrator credentials rather than provisioning narrowly scoped, individual accounts. ### Attack Path 1. The agency requests or receives administrative credentials during project onboarding. 2. The Skill generates a project-close email that includes those credentials with links and documentation. 3. The email is stored in the sender's and recipient's mailboxes and may be forwarded, archived, or synchronize ...[truncated 1070 chars]
Remediation
## Remediation Suggestions - Remove the instruction to place credentials in project-close emails or ordinary documentation. - Require secrets to be transferred through an approved password manager or an encrypted, access-controlled, expiring secret-sharing service. - Provision named, least-privilege accounts for agency personnel instead of sharing client administrator accounts. - Require multifactor authentication for all administrative and integration accounts. - Use separate communication channels when transmitting access instructions and secret-sharing links. - Rotate any credential that was previously shared through email, chat, or project documentation. - Revoke agency access promptly when delivery or the support period ends. - Maintain access logs and periodically review active users, API tokens, connected applications, and account permissions. - Store integration secrets in platform-native secret stores rather than spreadsheets, email, or shared folders. - Replace the affected prompt text with guidance such as: “Provide documentation and access-transfer instructions, but never include passwords, API keys, recovery codes, or other secrets in email. Transfer secrets through an approved encrypted secret-sharing mechanism and require rotation after handoff.”
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Confidence
93% confidence
Finding
The prompt explicitly normalizes discussing account credentials and admin access as part of kickoff logistics without any guidance to use secure sharing methods, least-privilege access, or secret management. In practice, users may respond by pasting passwords or broadly sharing privileged accounts in insecure channels, increasing the risk of credential theft, account compromise, and uncontrolled access.

Missing User Warnings

Medium
Confidence
96% confidence
Finding
The offboarding section tells the operator to provide credentials to the client, but omits any secure transfer, credential rotation, ownership transfer, or revocation guidance. That can leave shared secrets exposed after project completion, create lingering contractor access, and weaken the client's security posture long after delivery.

Static analysis

No suspicious patterns detected.