Back to skill

Security audit

Revops

Security checks for vulnerabilities and agentic risk

Overview

This RevOps skill is a non-executable guidance package; its CRM, Slack, and ad-tech automation examples are purpose-aligned but need normal privacy and admin review before use.

Before installing or using this skill to implement workflows, review any CRM, Slack/email, enrichment, and advertising automations with the relevant system owner. Limit notifications to need-to-know recipients, avoid sending full behavioral histories into chat or email when a CRM link will do, and verify consent, opt-outs, and jurisdictional rules before syncing contacts to ad platforms.

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
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Rogue AgentSelf-Modification, Session Persistence
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest description says to use the skill when the user wants help with revenue operations and also for 'anything involving the systems and processes that connect marketing to revenue.' That scope is broad and lacks clear exclusion boundaries or negative examples, increasing the chance of unintended invocation for general sales, marketing, or operations requests.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
85% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 227)May include surrounding context.

md
| Deal Size | Approval Required |
|-----------|-------------------|
| Standard pricing | Auto-approved |
| 10-20% discount | Sales manager |
| 20-40% discount | VP Sales |
| 40%+ discount or custom terms | Deal desk review |

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The eval explicitly expects the skill to trigger on 'casual phrasing' such as a vague complaint about pipeline issues, which broadens activation criteria beyond clearly bounded RevOps intents. Overly broad triggers can cause misrouting, where unrelated business or operational requests are incorrectly handled by this skill, reducing reliability and potentially surfacing inappropriate process guidance in the wrong context.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
80% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/automation-playbooks.md (reported line 30)May include surrounding context.

md
1. Send internal email to contact owner: "SLA warning: [Contact Name] has not been contacted"
2. If still no activity after 24 hours → send alert to sales manager
3. If still no activity after 48 hours → reassign contact owner via rotation
4. Create task for new owner: "Urgent: Contact [Contact Name] — reassigned due to SLA breach"
**Outcome:** No MQL goes unworked for more than 48 hours
**Notes:** Exclude contacts where last activity type is "Call" or "Meeting" (already engaged)

Session Persistence

Medium
Category
Rogue Agent
Confidence
80% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · references/automation-playbooks.md (reported line 73)May include surrounding context.

md
1. Send internal email to contact owner: "SLA warning: [Contact Name] has not been contacted"
2. If still no activity after 24 hours → send alert to sales manager
3. If still no activity after 48 hours → reassign contact owner via rotation
4. Create task for new owner: "Urgent: Contact [Contact Name] — reassigned due to SLA breach"
**Outcome:** No MQL goes unworked for more than 48 hours
**Notes:** Exclude contacts where last activity type is "Call" or "Meeting" (already engaged)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The playbook instructs sending meeting details, recent page views, and content downloads to internal recipients without any minimization, access-control, or privacy guidance. Even in a RevOps context, broad internal sharing of engagement and meeting-context data can expose sensitive personal or behavioral data to unnecessary audiences and increase compliance risk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The daily digest recipe distributes pages visited, content downloaded, and email engagement data to owners without any warning about sensitivity, need-to-know access, or lawful-use constraints. This can normalize mass internal redistribution of behavioral tracking data and create privacy, insider exposure, and regulatory issues.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

Posting enriched lead data to a Slack channel increases internal exposure of potentially sensitive personal and company information, especially when enrichment sources add firmographic or contact attributes. Without warnings about channel membership, retention, and minimization, the workflow can leak more data than is necessary for triage.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

Automatically adding leads to Facebook/Google retargeting audiences based on score thresholds raises consent, transparency, and ad-tech data-sharing concerns. In a RevOps playbook, this is contextually plausible but still risky because it operationalizes cross-tool advertising sync without mentioning legal basis, opt-out handling, or jurisdictional restrictions.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.