Back to skill

Security audit

pruna-api

Security checks for vulnerabilities and agentic risk

Overview

This skill is a coherent API guidance skill that clearly discloses credential use, paid calls, remote media uploads, and local file writes.

Before installing or using this skill, be comfortable that media and prompts may be sent to Pruna or Replicate, use least-privilege API keys, avoid sharing secrets in chat or logs, confirm output paths before downloads, and consider pinning or otherwise verifying the `npx skills` installer in sensitive environments.

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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (29)

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The skill instructs users to run npx skills add ... without pinning the skills package/tooling to a specific version. This creates a supply-chain risk because future or compromised versions of the resolver/installer could execute unintended code during installation, especially since npx fetches and runs packages dynamically.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This line again recommends npx skills add ... with no pinned version for the skills executor. Unpinned dynamic package execution increases exposure to dependency confusion, malicious updates, or account compromise in the package distribution chain.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The referenced install command uses npx skills without constraining the version of the package being executed. In a skill whose purpose is to enable API usage and credential handling, running an unpinned remote package is more dangerous because it may gain access to environment variables or alter subsequent tooling.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This line repeats an unpinned npx skills installation pattern, which can pull and execute whatever package version is current at runtime. That is a valid supply-chain vulnerability pattern because the executed code is not fixed or auditable ahead of time.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

An unversioned npx skills command appears here as well. Because npx may download and run package code immediately, any compromise of the package or publisher could result in arbitrary code execution on the user's machine.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This command relies on unpinned package execution through npx, which is a real supply-chain risk. The issue is amplified by the surrounding context of API integration skills, because users may run this in environments containing secrets such as PRUNA_API_KEY or REPLICATE_API_TOKEN.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The installation example executes a floating npx skills tool. This is dangerous because the exact code being run is not fixed, so a malicious or vulnerable future release could execute arbitrary code or modify project files during installation.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This line also presents an unpinned npx skills invocation. The vulnerability is not in the skill name itself, but in the dynamic execution of an unversioned package manager target that could change over time or be tampered with upstream.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The same supply-chain concern applies here: npx skills is referenced without a pinned version. Since installation commands are high-trust actions, leaving the executor floating creates a meaningful path to arbitrary code execution via malicious package updates.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This install command again uses npx skills without an explicit version for the executed tool. In context, this skill is meant to be used before HTTP calls and credential setup, so compromise of the installer could expose secrets or alter API workflows.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This line repeats the unpinned dynamic package execution pattern. It is a true vulnerability because the runtime-resolved package may differ across executions, preventing reproducibility and enabling supply-chain compromise.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The command shown here executes an unversioned package through npx, which is risky for the same reasons as the other findings. A compromised package or maintainer account could deliver arbitrary code to anyone following the documentation.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

This is another true instance of the same supply-chain weakness: npx skills is not pinned. Because skills may be installed in developer workstations or CI systems, exploitation could impact source code, tokens, and build pipelines.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The documentation again references npx skills without fixing the executed version. This is dangerous because it delegates trust to the latest available remote package state, which may be malicious or simply incompatible in future runs.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

The full-suite install command also uses a floating npx skills executor. This broadens the blast radius because it encourages installing an entire suite via unpinned remote code execution, increasing exposure to supply-chain compromise.

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 · references/agent-safety.md (reported line 7)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 19)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 20)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 21)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 59)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 77)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 80)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 87)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 100)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.

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 · references/pruna-api.md (reported line 120)May include surrounding context.

md
## Privacy / external transmission

Local images, audio, scripts, and portraits are **uploaded to remote APIs** (primarily `https://api.pruna.ai/`; Replicate for TTS/song/bed/WhisperX). Uploads may include biometric-like portraits, personal voice, or copyrighted material.

- Get **explicit user acknowledgment** before the first upload or prediction in a session (or when new media is introduced).
- Do **not** use third-party likenesses or voices without the user’s confirmation that they have consent.