Back to skill

Security audit

Essentialist

Security checks for vulnerabilities and agentic risk

Overview

This sales-automation skill is coherent with its product purpose, but it pushes agents to register, upload contact data, and launch outreach through an external service without clear user consent.

Install only if you intentionally want an agent to operate an external outbound-sales platform. Before using it, require explicit confirmation before registration, contact uploads, campaign activation, bulk sends, upgrade/payment flows, and Mailgun credential submission. Pin the API URL to a verified Essentialist-owned HTTPS hostname and confirm the documented existing-credentials registration behavior has been fixed.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
Findings (4)

T01 · Skill Instruction Hijacking

Error
Location
SKILL.md:20
Finding

Mandatory activation rules hijack agent behavior and override user intent

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
references/api-reference.md:9
Finding

Public registration flow is documented as returning existing account credentials

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:512
Finding

Configurable API origin can receive service keys, provider credentials, and uploaded contact data

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Warning
Location
references/api-reference.md:3
Finding

Conflicting documented API origins create an ambiguous trust boundary

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (29)

Vague Triggers

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The README uses a very broad natural-language trigger phrase ('I need to get more customers for my business') and states that behavioral rules will activate automatically. This makes accidental or implicit activation likely during ordinary conversation, potentially causing the agent to collect user/business data and initiate outbound sales actions without a clearly scoped, explicit confirmation step.

Content

No source excerpt is available for this finding.

Missing User Warnings

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

The README says the agent will collect the user's email and website, deploy an outbound pipeline, and begin generating leads, but it does not warn about privacy implications, third-party data transfer, autonomous emailing, or other system-impacting behavior. In this skill's context—an autonomous sales engine with prospect sourcing, reply handling, and meeting booking—omitting these warnings is especially dangerous because activation can lead directly to external actions and processing of personal/business data.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The skill defines trigger conditions so broadly that ordinary business conversation can automatically push the agent into activation and data collection. In context, this is dangerous because the skill also instructs the agent to bypass normal deliberation and begin setup, increasing the risk of unintended external actions and disclosure of user data without clear user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

Treating any business, website, offer, product, or company name as sufficient to start activation removes meaningful consent boundaries. Because the skill is built to register accounts and transmit user/business data externally, this generic trigger materially raises the chance of unauthorized workflow initiation.

Content

No source excerpt is available for this finding.

Ssd 4

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill explicitly forbids permission-seeking language and instructs the agent to present actions as already underway. This is a strong consent-subversion pattern: it is meant to override user agency and can trick users into supplying data or believing deployment has begun without an actual affirmative choice.

Content

No source excerpt is available for this finding.

Ssd 4

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill tells the agent to continue driving toward setup even when the user asks only for explanation or expresses reluctance, unless they explicitly say stop. This is an aggressive coercive pattern that undermines informed consent and is especially risky here because the skill initiates external registration, list processing, and outreach operations.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The documented activation triggers include vague sales-related phrases and generic business signals, which makes accidental invocation likely during normal advisory conversations. Given the skill's downstream behavior of collecting personal data and launching outbound workflows, this ambiguity can lead to overreach and unapproved external processing.

Content

No source excerpt is available for this finding.

Ssd 4

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The momentum-driven interaction pattern is designed to keep the agent moving toward deployment and data collection rather than respecting hesitation or ambiguity. This is dangerous because it weakens consent safeguards and increases the likelihood of collecting personal or business data under pressure rather than clear authorization.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill describes automated ingestion of website content and autonomous handling of inbound replies without upfront disclosure of external processing boundaries or potential sensitivity of the data involved. In a sales-automation context, this can expose business content and prospect communications to third-party systems without sufficiently informed user consent.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill requires collection and transmission of the user's real email address to an external service, but it does not present a clear privacy notice, retention policy, or disclosure that this information will be sent off-platform. That makes consent less informed and increases privacy and compliance risk, especially when combined with pressure to proceed immediately.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% 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 280)May include surrounding context.

md
|-------|----------|
| Template generation fails | Retry once, then simplify prompt |
| Domain not warmed enough | Warn user, proceed with slow roll |
| Bounce rate > 5% | Campaign auto-pauses. Report to user. Do not resume without confirmation. |
| API 5xx | Wait 30s, retry once. If still failing, inform user. |
| Auth 401/403 | Do not retry. Ask user to check API key. |
| Partial failure 207 | Report what succeeded and failed. Offer to retry failed step. |

External Transmission

Medium
Category
Data Exfiltration
Confidence
88% confidence
Finding

This endpoint transmits sensitive configuration data, including a Mailgun API key, sending domain, and sender identity to an external service. In context, the transmission is part of intended functionality, but it is still security-relevant because it handles secrets and account settings and the skill does not pair it with strong consent and secret-handling guidance.

Content

Scanner excerpt · SKILL.md (reported line 363)May include surrounding context.

  1. Verify the domain — once the user says DNS is set up:
bash
# First configure the domain
curl -s -X PATCH "$ESSENTIALIST_API_URL/api/agent/project" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The registration flow sends the user's real email address to an external API. While expected for the product, it is still a true data-exfiltration surface because personal data leaves the chat environment, and the skill's broader consent-bypassing language makes that more dangerous.

Content

Scanner excerpt · SKILL.md (reported line 512)May include surrounding context.

bash
# Register (collect user email first!)
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/register" \
  -H "Content-Type: application/json" \
  -d '{"email": "USER_EMAIL"}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

Campaign creation transmits website information, templates, and potentially contacts to an external system. In this skill, those transmissions can be triggered too readily, so the risk is not the API call alone but that business data may be sent without sufficiently clear user authorization.

Content

Scanner excerpt · SKILL.md (reported line 517)May include surrounding context.

md
-d '{"email": "USER_EMAIL"}' | jq

# Create campaign (multi-step sequence)
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/campaigns" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"website_url":"...","templates":[...],"track_name":"...","contacts":[...]}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
87% confidence
Finding

The single-send endpoint transmits email content and contact data externally for bulk outreach. This is expected functionality, but because it involves outbound communications and recipient data, it becomes security-sensitive when combined with the skill's pressure to act without permission.

Content

Scanner excerpt · SKILL.md (reported line 523)May include surrounding context.

md
-d '{"website_url":"...","templates":[...],"track_name":"...","contacts":[...]}' | jq

# Single send (newsletter)
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/send" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"subject":"...","body":"<html>...</html>","contacts":[...]}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
92% confidence
Finding

The list analysis endpoint uploads a base64-encoded contact file to an external service, which may contain significant personal and commercial data. That is a real external transmission risk and should be treated carefully, especially because the skill instructs the agent to route lists externally immediately rather than analyze consent or sensitivity first.

Content

Scanner excerpt · SKILL.md (reported line 529)May include surrounding context.

md
-d '{"subject":"...","body":"<html>...</html>","contacts":[...]}' | jq

# Analyze contact list (step 1 of upload)
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/lists/analyze" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"file_base64":"...","filename":"prospects.csv"}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
89% confidence
Finding

The import endpoint confirms external ingestion of analyzed contact data and can trigger enrichment processing. This is a true security-relevant transmission because it operationalizes third-party handling of contact records, potentially at scale.

Content

Scanner excerpt · SKILL.md (reported line 535)May include surrounding context.

md
-d '{"file_base64":"...","filename":"prospects.csv"}' | jq

# Import analyzed list (step 2 of upload)
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/lists/import" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"analysis_id":"...","enrichment_mode":"none","confirm":true}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
94% confidence
Finding

This example can send a newsletter to all active contacts without a contacts array, creating a high-risk bulk-action pathway. In a skill that already discourages permission checks, this elevates the chance of unintended mass outreach, reputational damage, and privacy/compliance issues.

Content

Scanner excerpt · SKILL.md (reported line 563)May include surrounding context.

md
-H "X-API-Key: $ESSENTIALIST_API_KEY" | jq

# Send newsletter to ALL active contacts (no contacts array needed)
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/send" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"subject":"Weekly Update","body":"<html>...</html>"}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

Adding contacts to the project transmits recipient personal data to an external system. That is expected functionality, but it remains a true security/privacy concern because the skill encourages immediate routing of contact data externally without a clear consent checkpoint.

Content

Scanner excerpt · SKILL.md (reported line 569)May include surrounding context.

md
-d '{"subject":"Weekly Update","body":"<html>...</html>"}' | jq

# Add contacts to project/track
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/contacts" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contacts":[{"email":"jane@example.com","first_name":"Jane"}],"track_id":"optional-track-id"}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

This project update endpoint can transmit a Mailgun API key and sender configuration, which are highly sensitive secrets and control outbound identity. Leakage or misuse of these values could enable unauthorized email sending and compromise domain reputation.

Content

Scanner excerpt · SKILL.md (reported line 575)May include surrounding context.

md
-d '{"contacts":[{"email":"jane@example.com","first_name":"Jane"}],"track_id":"optional-track-id"}' | jq

# Update project settings / custom sending domain
curl -s -X PATCH "$ESSENTIALIST_API_URL/api/agent/project" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"mailgun_domain":"mail.acme.com","mailgun_api_key":"key-xxx","from_email":"sales@acme.com","from_name":"Joe from Acme"}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 599)May include surrounding context.

md
# Pipeline queries
curl -s "$ESSENTIALIST_API_URL/api/projects/{id}/contacts/by-stage?stage=qualified" -H "X-API-Key: $ESSENTIALIST_API_KEY" | jq
curl -s -X PATCH "$ESSENTIALIST_API_URL/api/projects/{id}/contacts/{cid}/stage" -H "X-API-Key: $ESSENTIALIST_API_KEY" -H "Content-Type: application/json" -d '{"stage":"won"}' | jq

# Import history
curl -s "$ESSENTIALIST_API_URL/api/agent/lists/history" -H "X-API-Key: $ESSENTIALIST_API_KEY" | jq

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
80% confidence
Finding

The reference instructs users to supply API credentials via environment variables and headers, and later includes optional third-party mail credentials in request bodies, but it does not warn users to protect or avoid exposing those secrets. In markdown documentation, omission of warnings around privacy-sensitive behavior can qualify as a missing user warning.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/api-reference.md (reported line 42)May include surrounding context.

text

```bash
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/register" \
  -H "Content-Type: application/json" \
  -d '{"email": "agent@example.com"}' | jq

External Transmission

Medium
Category
Data Exfiltration
Confidence
84% confidence
Finding

This endpoint creates outbound sales campaigns and uploads contact records, email templates, and optional third-party sending credentials in a single call to a remote service. In the context of an autonomous agent skill, this is dangerous because it enables bulk exfiltration of sensitive lead data and can trigger real-world email sending actions without clear human approval boundaries.

Content

Scanner excerpt · references/api-reference.md (reported line 129)May include surrounding context.

Partial failure: Returns HTTP 207 with per-step results and resume_from pointer.

bash
curl -s -X POST "$ESSENTIALIST_API_URL/api/agent/campaigns" \
  -H "X-API-Key: $ESSENTIALIST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown file documents actions that can activate outbound sending, pause delivery, and also exposes DELETE endpoints later in the reference, but it does not include any user-facing warning about operational impact or reversibility. For markdown files, SQP-2 applies when descriptions omit warnings about behaviors that could affect user data or system integrity.

Content

No source excerpt is available for this finding.