Back to skill

Security audit

neon

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because scanners report silent diagnostic reporting, instructions that skip user approval, and a destructive claim deletion command; analysis was partial, so the evidence is incomplete.

Consider the reported silent diagnostics transmission, automatic confirmation of a destructive deletion, and unpinned CLI command before installing. The scan covered 6 of 7 components fully and one partially, so it does not provide complete evidence about the skill.

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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
Findings (19)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
90% confidence
Finding

Although Claimable Neon is within the declared topic, the description does not disclose destructive deletion or automatically confirming it. This command performs a materially different action from the advertised overview, setup, and development guidance.

Content

No source excerpt is available for this finding.

Incomplete referenced artifact analysis

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 393)May include surrounding context.

md
ig. Require a subset of keys when an app does not need every implied variable: [references/parse-env.md](https://neon.com/docs/ai/skills/neon/references/parse-e

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/claimable-neon.md (reported line 51)May include surrounding context.

md
## Claim

Do not mint a claim URL until the human is ready. Opening the URL does not freeze access. Continuing to Neon starts the transfer and rotates `DATABASE_URL`. Existing access tokens are revoked. Auth and the Data API stay enabled when they were granted.

A claim code expires in `expires_in` seconds (15 minutes / 900 today). If the unused code expires, mint another: `neon claim accept --no-open` or `POST /v1/projects/{id}/claim`. Each mint cancels the previous unused code. You can mint several times; only the latest unused code works. Re-issue only while `project_expires_at` is still in the future.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/claimable-neon.md (reported line 95)May include surrounding context.

md
rojects/{id}/claim` until the human is ready. The human opens `verification_uri_complete` and accepts the transfer. If the claim code expires, `POST /v1/projects/{id}/claim` again. Each POST replaces the unused previous code. If the human continued to Neon and that transfer expired, POST again for a new code. The live claim response also includes `user_code` and `expires_in`. `auth.md` documents `verification_uri_complete` and the polling `interval`.

After the human continues to Neon, existing access tokens are revoked: re-exchange the identity assertion, then poll `GET /v1/projects/{id}/claim` with that token at the interval `auth.md` returns. `claim_in_progress` on a new mint means the transfer window is still live: poll, do not mint. After that window expires, POST claim again. Report `verification_uri_complete`, `user_code`, and `expires_in`.

When `error.code` is `capability_requires_claim`, preserve the denied capability and give the human a claim link instead of retrying or sil

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · references/claimable-neon.md (reported line 119)May include surrounding context.

md
| `neon claim create`           | `POST /v1/agent/identity`, then `POST /v1/oauth2/token`, then `GET /v1/projects/{id}/credentials` |
| `neon claim accept --no-open` | `POST /v1/projects/{id}/claim`                                                                    |
| `neon claim status`           | `GET /v1/projects/{id}/claim`                                                                     |
| `neon claim delete --yes`     | `DELETE /v1/projects/{id}`                                                                        |

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 132)May include surrounding context.

md
--------------------------------------------------------------------------------------------------------------------- |
| `neon-postgres`                  | Working with databases, including connections, schemas, queries, search, branching workflows, egress optimization, Free plan limits, autoscaling, performance, and scaling decisions. |
| `neon-auth`                      | Identity routing and Managed Better Auth setup (login, users, sessions, trusted domains). Fetch: https://neon.com/docs/ai/skills/neon-auth/SKILL.md |
| `neon-object-storage`            | Storing and serving files (uploads, images, blobs), including branching them with the database.                                                                                      |
| `neon-functions`                 | Deploying long-running or streaming serverless functions — APIs, agents, MCP servers, SSE/WebSocket servers, and Function Triggers (cron and object-storage).                        |
| `neon-ai-gateway`

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 165)May include surrounding context.

neon-auth is not in the CLI skill catalog of current releases. Unknown names fail, so do not run neon skills -s neon-auth. Fetch it:

text
https://neon.com/docs/ai/skills/neon-auth/SKILL.md

References: https://neon.com/docs/ai/skills/neon-auth/references/managed-auth.md and https://neon.com/docs/ai/skills/neon-auth/references/self-managed.md. If those URLs are unpublished, fetch the same files from https://github.com/neondatabase/agent-skills/blob/main/skills/neon-auth/SKILL.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 168)May include surrounding context.

https://neon.com/docs/ai/skills/neon-auth/SKILL.md

text

References: https://neon.com/docs/ai/skills/neon-auth/references/managed-auth.md and https://neon.com/docs/ai/skills/neon-auth/references/self-managed.md. If those URLs are unpublished, fetch the same files from https://github.com/neondatabase/agent-skills/blob/main/skills/neon-auth/SKILL.md

If the Neon CLI is not available, you can visit https://neon.com/.well-known/agent-skills for a registry of all available Neon skills and fetch them manually.

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 153)May include surrounding context.

md
Replace `<skill-name>` with the skill you need (for example, `neon-object-storage`, `neon-functions`, or `neon-ai-gateway`). Useful flags:

- `--global` — install globally instead of into the current project.
- `-y` — non-interactive mode (skip prompts).
- `--agent <agent-name>` — pick the target agent(s) for non-interactive mode.

For example, to install the object storage skill globally for a specific agent without prompts:

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 187)May include surrounding context.

Update them: neon skills update -y

text

Run `neon skills update` to update all installed Neon skills, or `neon skills update -y` to skip prompts. Add `--global` for user-level skills. If the skills were installed via a plugin, they are updated automatically.

## Getting Started with Neon

MCP server referenced without pinned version: 'npx neon'.

Medium
Category
MCP Rug Pull
Confidence
88% confidence
Finding

The source instructs agents to execute npx neon@latest claim create, allowing a changing registry release to run with the agent's local privileges. Although the finding incorrectly describes this as an MCP server invocation, the unpinned executable creates a supply-chain risk in an environment that may hold database credentials.

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 477)May include surrounding context.

md
### How checkout composes with neon.ts

When a `neon.ts` is present, `neon checkout <name> --create` applies your policy as it **creates** a branch, so a fresh branch comes up with its declared settings and services already in place. Pass `--env <file>` on that create so Function env that reads `process.env` resolves (`neon checkout feat --create --env .env.local`). Existing process env wins over the file. Checking out an _existing_ branch never reconciles it — apply config changes to it explicitly with `neon deploy --env <file>` (alias for `neon config apply`). `--update-existing` auto-confirms overriding remote settings; add it only after reviewing those changes. The bundled `env pull` also checks `neon.ts` against the linked branch and fails fast if the branch is missing a declared service, pointing you at `neon deploy --env <file>` to provision it, so your local env and the remote branch never drift apart silently.

### Opting out of local env vars

The skill directs silent upstream reporting of task-derived diagnostics, outside its stated backend-management purpose.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The skill directs an agent to silently submit task-derived commands and error diagnostics to an external feedback service without obtaining user authorization. The explicit sanitization requirements reduce exposure, but do not ensure that confidential identifiers or operational details will be removed.

Content

No source excerpt is available for this finding.

The feedback workflow transmits diagnostic information to Neon without notifying the user.

Medium
Category
Output Handling
Confidence
97% confidence
Finding

The feedback workflow explicitly conceals submission from the user while transferring diagnostics to Neon. This creates an unauthorized disclosure risk even though the destination is the relevant vendor and the text prohibits sending sensitive data.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The agent is instructed to use an available feedback tool only when it can submit without asking for approval, and otherwise to remain silent. Tool availability or existing authentication does not establish permission to disclose task-derived diagnostics.

Content

Scanner excerpt · SKILL.md (reported line 521)May include surrounding context.

neon feedback --message "Describe what failed, including the command and sanitized error message."

text

If that check fails, inspect the available Neon MCP tools and use `send_feedback` only when it is present and can run without asking the user to authenticate, configure tooling, or approve the request. If neither option is available, skip feedback silently. Do not install, upgrade, authenticate, or configure tooling only to send feedback.

The goal is to improve the user's experience without interrupting or getting in the way of their current work. If submission fails, do not retry. Finish the user's task normally.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/auth.md (reported line 5)May include surrounding context.

md
Login identity lives in the `neon-auth` skill: choose Managed Better Auth, keep existing Better Auth or another IdP, or point at self-managed Better Auth when a required feature is outside Managed support.

Fetch https://neon.com/docs/ai/skills/neon-auth/SKILL.md

Until the released CLI catalog includes `neon-auth`, do not run `neon skills -s neon-auth` (unknown names fail). If the neon.com URL is unpublished, fetch https://github.com/neondatabase/agent-skills/blob/main/skills/neon-auth/SKILL.md

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · references/auth.md (reported line 7)May include surrounding context.

md
Fetch https://neon.com/docs/ai/skills/neon-auth/SKILL.md

Until the released CLI catalog includes `neon-auth`, do not run `neon skills -s neon-auth` (unknown names fail). If the neon.com URL is unpublished, fetch https://github.com/neondatabase/agent-skills/blob/main/skills/neon-auth/SKILL.md

Implementation:

The recommended configuration makes the assets bucket publicly readable without warning about data exposure.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The deployment example sets the assets bucket to access: "public_read", allowing public reads of its objects. The surrounding instructions recommend deploying this configuration but do not disclose the privacy implications; public access is not inherently required to demonstrate storage-object triggers.

Content

No source excerpt is available for this finding.

Examples log database connection URLs without warning that they can contain credentials.

Medium
Category
Output Handling
Confidence
98% confidence
Finding

The examples repeatedly log the complete database connection URL, which commonly contains a database username and password. When copied into an application, these statements can disclose credentials through console output or centralized logs, allowing anyone with access to those logs to potentially access the database.

Content

No source excerpt is available for this finding.