Back to skill

Security audit

customer-support-memory

Security checks for vulnerabilities and agentic risk

Overview

This skill is purpose-built for external customer-support memory, but it stores and recalls potentially sensitive customer data persistently without enough privacy, scoping, or deletion guidance.

Review before installing. Use this only if your organization has approved BlueColumn and the Supabase endpoint as a data processor, and configure agents to store only minimized, redacted summaries. Do not store secrets, payment data, regulated data, or unnecessary PII. Require identity and tenant checks before recall, keep API keys in a secret manager, and confirm retention and deletion procedures before enabling routine use.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Findings (1)

other

Warning
Location
SKILL.md:9
Finding

Uncontrolled External Storage of Customer Support Data

Content
View full analysis
" \ -H "Content-Type: application/json" \ -d '{"text": "Ticket #4412: Jane reported rate limiting on Developer plan. Escalated to billing.", "title": "customer-support-memory - note"}' ``` ``` ```markdown ## Workflow 1. On new context, first recall: `What is the history of Jane's support tickets?` 2. Use the answer to personalize the response 3. After the interaction, store the summary via `/agent-remember` ``` ### Technical Analysis The skill instructs the agent to retrieve an API credential from `TOOLS.md` or the platform secret store and use it to transmit customer-support information to an externally hosted Supabase endpoint. It also makes external storage part of the standard post-interaction workflow. Customer-support conversations can contain personally identifiable information, account details, confidential correspondence, authentication material, billing information, or regulated data. The instructions do not require user consent, data classification, redaction, secret detection, data minimization, tenant validation, or an approved-domain check before transmission. They also do not document retention, deletion, access-control, or data-processing guarantees for the persistent external memory. The external endpoint uses a `supabase.co` hostname rather than the advertised `bluecolumn.ai` domain. This is not inherently malicious, but it reduces endpoint transparency and makes independent verification of the recipient and data-processing boundary more important. ### Attack Path 1. A customer includes ...[truncated 1553 chars]
Remediation
View remediation
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 (4)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill advertises storing and recalling customer-support context in an external persistent memory service but does not warn users that potentially sensitive customer data will be transmitted off-platform and retained. In a support setting, stored ticket history may include PII, account details, and internal case notes, so omission of a privacy/transmission warning materially increases disclosure risk.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
94% confidence
Finding

The skill includes a concrete example that sends support-ticket content to an external endpoint using an authorization bearer token. External transmission is expected for this integration, but it is still security-relevant because the example payload contains customer-specific support details and the document provides no safeguards around sensitive data handling, consent, or minimization.

Content

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

Store

bash
curl -X POST .../agent-remember \
  -H "Authorization: Bearer <key>" \
  -H "Content-Type: application/json" \
  -d '{"text": "Ticket #4412: Jane reported rate limiting on Developer plan. Escalated to billing.", "title": "customer-support-memory - note"}'

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The workflow explicitly tells the agent to store post-interaction summaries via /agent-remember without any notice about persistence, retention, or privacy consequences. Because support summaries often condense sensitive facts about a customer's account and prior issues, this creates a real risk of silent long-term storage and later misuse or over-disclosure.

Content

No source excerpt is available for this finding.

Ssd 3

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The skill instructs the agent to recall a named customer's prior support history and use it to personalize future responses, which introduces a cross-session data retention and disclosure risk. In customer support contexts, this can cause unnecessary accumulation of personal history and increase the chance that unrelated future interactions expose prior sensitive details to the wrong user or context.

Content

No source excerpt is available for this finding.