Most shell-execution and reconnaissance alerts are false positives caused by Markdown documentation. Confirmed risks include runtime execution of a mutable npm package, plaintext storage and use of a full-access credential, and high-impact financial and communication actions without enforced safeguards. Attachment, contact backup, and automatic webhook credential discovery expand the data exposure surface.
account: new ATXPAccount(process.env.ATXP_CONNECTION),
The SDK example uses the sensitive ATXP_CONNECTION credential to authenticate a client. Exposure of this value grants the documented wallet and identity access.
- **Skill env config:** Set `skills.entries.atxp.env.ATXP_CONNECTION` in `openclaw.json`
The documentation recommends placing ATXP_CONNECTION in runtime configuration or an environment file. Such storage can expose the credential if file permissions or configuration access are weak.
- **Environment file:** Add `ATXP_CONNECTION=<value>` to the `.env` file loaded by the agent runtime
The documentation recommends placing ATXP_CONNECTION in runtime configuration or an environment file. Such storage can expose the credential if file permissions or configuration access are weak.
The skill instructs users to invoke npx with the mutable @latest tag, which downloads and executes the package and dependencies at runtime. A compromised future release can execute with the agent host permissions.
The runtime download, local subprocess execution, and widespread use of @latest are explicitly documented.
ATXP_CONNECTION grants full wallet and identity access and is stored automatically in a plaintext configuration file. The stated user-read-only permission is advisory documentation, not an enforced secret-store control.
The documentation explicitly identifies the credential as full access and the file as plaintext.
The skill can fund accounts, spend credits, send email or SMS, and place calls, while its human approval and spending caps are recommendations that the skill does not enforce. An autonomous caller can therefore initiate high-impact actions without a technical gate.
The documented commands perform paid and outbound actions, and the text expressly says the skill does not enforce limits.
The notifications setup states that it auto-discovers a webhook URL and authentication token in OpenClaw instances. Automatic discovery expands the credential and endpoint exposure surface without a documented confirmation step.
The documentation explicitly describes automatic discovery of both a webhook URL and auth token.
Email attachment commands read local files for transmission, and contact push backs up contact data to a server. The documentation does not describe enforced path allowlists, recipient controls, or cloud-backup consent gates.
The file-reading, base64 encoding, outbound attachment, and contact backup behavior are explicitly documented.
Capability review items (25)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
compatibility: Requires Node.js >=18 and npx. Network access to *.atxp.ai (HTTPS only). Writes to ~/
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
description: Authentication token for the ATXP API. Created by `npx atxp@latest login` or `npx atxp@
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
- Filesystem writes are limited to `~/.atxp/config` (auth credential only) and the standard npm cach
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The `ATXP_CONNECTION` credential is created automatically by `npx atxp@latest login` or `npx atxp@la
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
**Important:** `ATXP_CONNECTION` is a sensitive credential. Never pass it as a CLI argument, echo it
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
Local contacts database for resolving names to phone numbers and emails. Stored in `~/.atxp/contacts
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
compatibility: Requires Node.js >=18 and npx. Network access to *.atxp.ai (HTTPS only). Writes to ~/
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
description: Authentication token for the ATXP API. Created by `npx atxp@latest login` or `npx atxp@
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
- Filesystem writes are limited to `~/.atxp/config` (auth credential only) and the standard npm cach
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
The `ATXP_CONNECTION` credential is created automatically by `npx atxp@latest login` or `npx atxp@la
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
**Important:** `ATXP_CONNECTION` is a sensitive credential. Never pass it as a CLI argument, echo it
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
Local contacts database for resolving names to phone numbers and emails. Stored in `~/.atxp/contacts
The documentation declares local storage or automatic reading of an ATXP credential or contact data under the user home directory. Compromise of that data can expose account access or personal contact information.
This skill executes code downloaded from the npm registry at runtime via `npx`. This is an inherent
The documentation states that npx downloads and runs the atxp package at runtime, and the documented commands use the mutable @latest tag. This creates a supply-chain execution risk if a release or dependency is compromised.
| **Version pinning** | All commands use `npx atxp@latest` for convenience. For stricter supply-chai
The documentation states that npx downloads and runs the atxp package at runtime, and the documented commands use the mutable @latest tag. This creates a supply-chain execution risk if a release or dependency is compromised.
| **npm runtime** | `npx atxp@latest` downloads the `atxp` package from the npm registry and caches
The documentation states that npx downloads and runs the atxp package at runtime, and the documented commands use the mutable @latest tag. This creates a supply-chain execution risk if a release or dependency is compromised.
account: new ATXPAccount(process.env.ATXP_CONNECTION),
The SDK example reads ATXP_CONNECTION from the process environment. This is a real credential access point, although it is a conventional integration pattern.
account: new ATXPAccount(process.env.ATXP_CONNECTION),
The SDK example reads ATXP_CONNECTION from the process environment. This is a real credential access point, although it is a conventional integration pattern.
Many static backtick matches are Markdown examples, but several document real npx, MCP, email, phone, and paid API actions. The main risks are runtime npm execution, external ATXP network calls, wallet spending, outbound communications, and local storage of ATXP_CONNECTION and contacts. No prompt injection attempt was found in SKILL.md.
Funded Agent Can Spend Credits and Contact External Parties
The skill enables an agent to use a funded wallet, paid search, media, LLM tools, email, SMS, and voice calls. Without approval controls, compromised prompts or automation errors could spend credits or send data to arbitrary recipients.
The cited lines explicitly label paid commands and warn that email, SMS, and calls are high-risk actions requiring approval.
Plain-Text Credential Grants Wallet and Identity Access
The ATXP_CONNECTION token grants wallet and identity access and is stored in a local config file or environment configuration. Exposure of this token could allow unauthorized use of the agent account.
The skill states the token is a sensitive secret with full wallet and identity access and documents local and environment-based storage options.
The skill relies on npx atxp@latest, which downloads and executes the npm package and dependencies at runtime. Using latest can change behavior between invocations unless operators pin and verify a version.
The skill explicitly states that npx downloads ATXP from npm and recommends exact version pinning as a supply-chain control.
The `email send` and `phone send-sms` commands can transmit data to arbitrary addresses. To prevent
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
- Treat `email send`, `email reply`, `phone send-sms`, and `phone call` as high-risk actions requiri
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
| `npx atxp@latest email reply <messageId> --body <body>` | $0.01 | Reply to email ⚠️ EXFILTRATION R
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
**Sending attachments:** Use the `--attach` flag (repeatable) with `email send` or `email reply` to
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
Register a phone number to send/receive SMS and make/receive voice calls. The phone command is async
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
| `npx atxp@latest phone call --to <number> --instruction <text>` | $0.10 | Make a voice call |
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
This line documents outbound communication or attachment capabilities that an agent can invoke through ATXP, creating data-exfiltration and abuse risk.
The skill is a thin reference for ATXP CLI and MCP services, and most markdown backtick detections are not Ruby execution. Several findings remain valid because the instructions call external npx commands, source a hidden config file, read an ATXP connection token, and contact ATXP network services.
account: new ATXPAccount(process.env.ATXP_CONNECTION),
The exact code reads an environment variable rather than an environment file, but `ATXP_CONNECTION` is used as account credential material. Keeping the sensitive finding is appropriate because the secret is passed into the ATXP client.
The skill describes paid API tools and includes commands for generation services plus a PaaS deployment reference. Automated use could spend credits or create cloud resources without clear user confirmation.
The file explicitly says the tools are paid and lists media generation commands plus serverless deployment capabilities. This is a business-logic risk separate from the raw command and network patterns.
Capability review items (13)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
Line 18 instructs `source ~/.atxp/config`, which reads and executes shell content from a hidden home-directory file. That can load secrets and run arbitrary shell statements.
description: Access ATXP paid API tools for web search, AI image generation, music creation, video g
The cited description explicitly requires `npx atxp login`, which executes an external npm package during authentication. The markdown backticks are not Ruby execution, but the command instruction is real.
The bash block includes `npx atxp login` and `source ~/.atxp/config`. Running an npm package and sourcing a shell config are real external command actions.
The finding starts at a code-fence close, but its cited range reaches the command table that begins with `npx atxp search <query>`. This is an external CLI command instruction, so it is kept.
| `npx atxp search <query>` | Real-time web search |
Line 25 documents `npx atxp search <query>`, which executes the ATXP CLI and sends a query to an external service. This is intentional but security-relevant command execution.
| `npx atxp image <prompt>` | AI image generation |
Line 26 documents `npx atxp image <prompt>`, which runs an external CLI and transmits prompt content for image generation. This is intentional but still an external command risk.
| `npx atxp music <prompt>` | AI music generation |
Line 27 documents `npx atxp music <prompt>`, which runs an external CLI and sends user prompt data to a paid service. This is a real external command pathway.
| `npx atxp video <prompt>` | AI video generation |
Line 28 documents `npx atxp video <prompt>`, which runs an external CLI and sends prompt content to a video generation service. This is expected but security-relevant.
Line 29 documents `npx atxp x <query>`, which runs the ATXP CLI and sends search terms to an external X search service. This is real external command use.
Line 18 accesses `~/.atxp/config`, a hidden user configuration path likely to contain authentication material. The access is intentional but security-sensitive.
The TypeScript sample hardcodes `https://search.mcp.atxp.ai` as the MCP server. This is intentional, but it creates outbound network communication to a third-party service.
account: new ATXPAccount(process.env.ATXP_CONNECTION),
Line 48 reads `process.env.ATXP_CONNECTION` for account authentication. This is expected for the client, but it uses an environment value that may contain a secret token.
account: new ATXPAccount(process.env.ATXP_CONNECTION),
Line 48 accesses the environment object to retrieve `ATXP_CONNECTION`. The value appears to be an authentication connection string and should be treated as sensitive.
Static analysis correctly identified external command, network, filesystem, and environment access patterns, but most backtick and weak-crypto alerts are Markdown formatting false positives. The skill still carries medium risk because it instructs agents to run `npx atxp`, source a hidden credential file, and call paid external MCP services.
1
Files scanned
67
Lines analyzed
7
Review items
2
False positives ignored
Capability review items (3)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
The skill instructs users or agents to run `npx atxp login` and later use `npx atxp` commands for search and media generation. This is legitimate for the stated purpose, but it executes third-party CLI code and can incur paid API actions.
The command examples are explicit and central to the skill workflow. They are not evidence of command injection, but they do require external command execution.
The skill asks users to inspect `$ATXP_CONNECTION`, source `~/.atxp/config`, and pass `process.env.ATXP_CONNECTION` into `ATXPAccount`. This is expected authentication behavior, but it exposes a credential-bearing environment variable to the agent runtime.
The credential path and environment variable usage are directly shown. The risk is clear, although the use appears aligned with authentication rather than malicious collection.
The skill directs calls to ATXP-hosted MCP servers for search, image, music, video, X/Twitter search, and PaaS tools. These network services are part of the advertised function, but requests may send user prompts and consume paid resources.
The external endpoint and server list are explicit. The skill purpose makes network use legitimate, but the paid service and data-sharing implications remain material.
Static false positives ignored (2)
These static matches were dismissed by semantic review or matched schema-only tokens, so they are shown for transparency but do not drive the quality score.
Static Backtick And Weak Crypto Alerts Are False Positives
The analyzer flagged many Markdown backticks as Ruby or shell backtick execution and weak cryptography. The reviewed lines are frontmatter text, Markdown tables, fenced examples, or headings, with no executable Ruby code or cryptographic algorithm use found.
The file is Markdown and the flagged characters are formatting delimiters around commands and server names. No evidence of Ruby execution or cryptographic implementation was found.
The reviewed skill text does not contain instructions to ignore prior instructions, override system policy, skip security analysis, or claim pre-approval.
The full skill file was reviewed and no suspicious authority-changing or analysis-skipping language was present.
All 36 static analysis findings have been evaluated and dismissed as false positives. The detected 'external_commands' are documentation examples (shell commands in backticks), not actual code execution. Network URLs point to legitimate ATXP API endpoints. Filesystem and environment access patterns are for standard configuration management. This skill provides documentation for accessing a legitimate paid API service.
1
Files scanned
67
Lines analyzed
0
Review items
0
False positives ignored
No confirmed security findings were recorded for this completed audit.