Back to skill

Security audit

Prisma Access All-in-One

Security checks for vulnerabilities and agentic risk

Overview

The skill is coherent for Prisma Access administration, but it can change, delete, and push security configuration using tenant credentials with limited explicit safeguards.

Review this skill before installing in production. Use a least-privileged SCM service account, require explicit confirmation for every create, update, delete, migration, and push, prefer dry runs with impact previews, and avoid the documented curl secret pattern on shared machines.

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
SKILL.md:167
Finding

SCM Client Secret Exposed Through Process Arguments

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 167–172
Vulnerability Type: OAuth client-secret exposure through command-line arguments
Risk Level: Medium

Vulnerable Code

bash
TOKEN=$(curl -s -X POST "https://auth.apps.paloaltonetworks.com/am/oauth2/access_token" \
  -H "Content-Type: application/x-www-form-urlencoded" \
  -d "grant_type=client_credentials" \
  -d "client_id=${SCM_CLIENT_ID}" \
  -d "client_secret=${SCM_CLIENT_SECRET}" \
  -d "scope=tsg_id:${SCM_TSG_ID}" | jq -r '.access_token')

Technical Analysis

The shell expands ${SCM_CLIENT_SECRET} before starting curl, placing the plaintext OAuth client secret in the process argument vector. While the request is running, local process-inspection mechanisms may expose the expanded argument to another user or process with sufficient observation privileges.

Transport encryption protects the request in transit but does not mitigate this local disclosure because the secret is exposed before TLS transmission. The issue does not represent deliberate exfiltration: the destination is the documented Palo Alto Networks authentication endpoint. Nevertheless, the documented authentication pattern handles sensitive credentials insecurely.

Attack Path

  1. An attacker first obtains local access with permission to inspect process arguments, such as through process-monitoring interfaces or administrative tooling.
  2. The operator invokes the Skill's documented SCM authentication command.
  3. The shell expands ${SCM_CLIENT_SECRET} into curl's command-line argument list.
  4. The attacker monitors the short-lived curl process and captures the client ID, client secret, and tenant scope from its arguments.
  5. The attacker submits those credentials to the legitimate SCM OAuth endpoint to obtain an access token.
  6. The attacker uses that token to access SCM APIs within the permissions assigned to the compromised service account.

Impact Assessment

Successful exploitation c ...[truncated 638 chars]

Remediation
View remediation

Remediation Suggestions

  1. Supply the form data through standard input so the secret is not included in the process argument vector. For example:
bash
TOKEN=$(
  printf '%s' \
    "grant_type=client_credentials&client_id=${SCM_CLIENT_ID}&client_secret=${SCM_CLIENT_SECRET}&scope=tsg_id:${SCM_TSG_ID}" |
  curl --silent --show-error --fail \
    -X POST \
    "https://auth.apps.paloaltonetworks.com/am/oauth2/access_token" \
    -H "Content-Type: application/x-www-form-urlencoded" \
    --data-binary @- |
  jq -er '.access_token'
)
  1. Prefer a dedicated secret manager or protected credentials file with restrictive permissions when supported by the execution environment.
  2. Disable shell tracing around authentication operations because set -x may reveal expanded secrets even when standard input is used.
  3. Do not log request bodies, environment variables, OAuth responses, or access tokens.
  4. Use a least-privileged SCM service account and restrict its roles to the minimum resources and operations required.
  5. Rotate the client secret if the vulnerable command has been used on systems where untrusted users could inspect processes.
  6. Add explicit error handling with curl --fail and jq -e so authentication failures do not result in malformed or inadvertently logged token values.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% confidence
Finding

The skill explicitly supports arbitrary DELETE operations against SCM resources, creating a clear path for destructive changes if the agent misinterprets user intent, is prompted ambiguously, or is influenced by injected content. In a multi-tenant configuration-management context, mistaken or unauthorized deletions can disrupt security policy, connectivity, and production access at scale.

Content

Scanner excerpt · SKILL.md (reported line 200)May include surrounding context.

md
**Update**: `PUT /sse/config/v1/{resource}/{id}` with JSON body

**Delete**: `DELETE /sse/config/v1/{resource}/{id}`

**Push Config**: `POST /sse/config/v1/config-versions/candidate:push`

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The skill maps very broad synonyms like "list", "create", "update", and "delete" directly into API-operation mode based on free-form user arguments. This increases the chance of unintended invocation of sensitive workflows, including destructive actions, from ambiguous prompts or prompt-injection-style phrasing in surrounding context.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% 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 60)May include surrounding context.

Always output as SCM API-compatible JSON:

text
POST https://api.sase.paloaltonetworks.com/sse/config/v1/{resource}?folder={folder}

Include the JSON payload, API endpoint, required folder parameter, and any query parameters.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The API operations section documents create, update, delete, and push capabilities but does not present a prominent user-facing warning before destructive or high-impact operations. In an agent skill, this omission can lead to unsafe execution of tenant-changing actions without sufficiently explicit informed consent, especially when combined with broad command routing.

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 · SKILL.md (reported line 178)May include surrounding context.

Authentication

bash
TOKEN=$(curl -s -X POST "https://auth.apps.paloaltonetworks.com/am/oauth2/access_token" \
  -H "Content-Type: application/x-www-form-urlencoded" \
  -d "grant_type=client_credentials" \
  -d "client_id=${SCM_CLIENT_ID}" \