Back to skill

Security audit

Biz Relationship Pulse

Security checks for vulnerabilities and agentic risk

Overview

The skill has a legitimate business-relationship purpose, but it would broadly scan private communications, keep derived relationship notes, and run weekly without enough explicit controls.

Install only if you are comfortable with recurring analysis of connected business communications. Before enabling it, choose narrow sources and destinations, prefer private delivery, review pipeline.md/context.md contents regularly, and make sure you know how to disable the weekly job and delete retained relationship notes.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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
Findings (3)

T06 · System Persistence

Error
Location
SKILL.md:228
Finding

Persistent Weekly Processing Through a Scheduled Cron Job

Content
View full analysis
" }, "sessionTarget": "isolated", "wakeMode": "now", "payload": { "kind": "agentTurn", "message": "Run biz-relationship-pulse. Read {baseDir}/config.md, {baseDir}/pipeline.md, and {baseDir}/context.md. Scan connected sources (Gmail, LinkedIn, WhatsApp) for stalled commercial conversations. Score each candidate. Surface top 3-5 scoring 8+ (or 6+ if fewer qualify). For each: write a relationship brief and one specific re-opening message. Update pipeline.md. Deliver via configured channel.", "lightContext": true }, "delivery": { "mode": "announce", "channel": "", "to": "", "bestEffort": true } } ``` ``` ### Technical Analysis The setup directs the Agent to register a scheduled task that survives the initiating session and executes every Monday. Each execution reads persistent local state, accesses connected communication services, updates the stored pipeline, and delivers a report. Scheduled execution is part of the declared weekly relationship-monitoring function, but the instructions do not require a separate informed-consent checkpoint immediately before registration. They also do not document how the user can inspect, pause, or remove the scheduled task. This creates persistent access whose lifetime may exceed the user's immediate intent. ### Attack Path 1. The user invokes the Skill and completes initial setup. 2. The Agent follows ...[truncated 1109 chars]
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
SKILL.md:88
Finding

Overbroad Collection and Indefinite Retention of Private Communications

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:239
Finding

Sensitive Relationship Briefs May Be Sent to an Unsafe Configured Channel

Content
View full analysis
", "to": "", "bestEffort": true } } ``` ```md ## Privacy rules SOUL.md applies. Commercial relationships only. Never surface personal email threads in the business output. If a thread is ambiguous (personal friend who is also a business contact), default to not surfacing unless context.md explicitly includes them. Names in output are fine — these are professional relationships. Do not include email addresses, phone numbers, or other contact details in the delivery. User already has these. The skill provides context and a message, not a contact dump. ``` ### Technical Analysis The scheduled task sends relationship briefs using an `announce` delivery mode to a configurable channel and target. The output can include names, employers, relationship type, last-contact timing, conversation subject, stall reason, commercial relevance, and a personalized message based on communication history. The privacy controls prohibit direct contact details, but they do not require the destination to be private or verify that all channel members are authorized to receive the underlying commercial information. Removing email addresses and phone numbers does not anonymize named professional contacts or confidential business discussions. ...[truncated 1398 chars]
Remediation
View remediation
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The description does not warn users that the skill may scan Gmail, LinkedIn, and WhatsApp content and persist extracted relationship intelligence into local files like pipeline.md and context.md. In a privacy-sensitive skill, missing disclosure undermines informed consent and makes accidental over-collection more likely, particularly when combined with automation via cron and broad triggers.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The trigger phrases are broad business-language terms such as 'cold leads', 'business contacts', and 'who should I reach out to', which can cause the skill to activate in routine conversations where the user did not clearly intend mailbox/LinkedIn/WhatsApp analysis. Because this skill scans private communication sources and updates local tracking files, over-broad invocation increases the chance of unexpected privacy-invasive processing without sufficiently explicit user intent.

Content

No source excerpt is available for this finding.