Back to skill

Security audit

Piper TTS

Security checks for vulnerabilities and agentic risk

Overview

This skill describes a useful local text-to-speech workflow, but it tells the agent to run missing setup and speech scripts that cannot be reviewed.

Review this before installing. Only use it when you explicitly want text-to-speech, and do not run the referenced scripts unless you have the actual files from a trusted source and have inspected them. Prefer a version that includes auditable setup scripts, pinned package/model versions, checksums, and clear disclosure of setup network access.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:18
Finding

Unpinned and Unverifiable Third-Party Dependency Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 18-27
Vulnerability Type: T08: Insecure Dependencies
Risk Level: Medium

markdown
## Setup

If Piper is not installed, run the setup script:

```bash
scripts/setup-piper.sh

This installs piper-tts via pip and downloads a default voice (en_US-kusal-medium).

text

### Technical Analysis

The Skill instructs users to execute `scripts/setup-piper.sh`, which purportedly installs the `piper-tts` Python package and downloads a voice model. However, that script is absent from the audited project, whose only file is `SKILL.md`. Consequently, its package source, version constraints, download URLs, shell commands, and integrity-validation behavior cannot be reviewed.

The documentation does not specify an exact dependency version, package hash, model version, checksum, or signature. Installing a mutable package release and downloading unverified model content creates a supply-chain exposure: the retrieved artifacts can differ from those intended or previously reviewed. The missing script also makes the documented setup process unverifiable and could cause an operator or agent to execute a same-named script obtained from an unintended location.

This finding does not establish that `piper-tts` or the referenced voice model is malicious. The risk arises from the undocumented, unpinned, and unverifiable installation process.

### Attack Path

1. A user or agent loads the Skill and follows its setup instructions.
2. The environment supplies the missing `scripts/setup-piper.sh` from another source, or a later package revision adds it without equivalent review.
3. The script installs the current package release and downloads model content without documented version or integrity controls.
4. A compromised package release, substituted download, dependency-confusion package, or malicious replacement script is retrieved.
5. Package installation logic or shell 
...[truncated 907 chars]
Remediation
View remediation

Remediation Suggestions

  1. Include scripts/setup-piper.sh and scripts/piper-speak.sh in the project so their behavior can be audited.
  2. Pin piper-tts and all transitive dependencies to reviewed versions.
  3. Use a lock file or hash-verified installation, such as pip requirements with --require-hashes.
  4. Pin voice models to immutable revisions and verify a published cryptographic checksum or signature before use.
  5. Download only from explicitly documented HTTPS endpoints and reject redirects to untrusted origins.
  6. Avoid sudo, system-wide package installation, and other privilege elevation; install into an isolated virtual environment under a least-privileged account.
  7. Fail securely if integrity verification fails, and remove incomplete downloads.
  8. Document every network destination and clarify that external access is required during setup despite operation being local afterward.
  9. Ensure script paths are resolved relative to a trusted, verified Skill directory rather than the current working directory.
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Vague Triggers

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The activation guidance is incomplete because 'Use when' is left unfinished and does not clearly define invocation boundaries, prerequisites, or consent requirements. Ambiguous activation increases the chance that the agent applies this skill in the wrong context and executes setup or speech-generation steps without sufficient user intent.

Content

No source excerpt is available for this finding.

Vague Triggers

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The core capability section repeats broad trigger terms, reinforcing a collision-prone activation pattern across the document. Repetition of generic triggers raises the probability that orchestration logic or an LLM-based router selects this skill for unrelated prompts, leading to unintended command execution or media handling.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The activation guidance is incomplete because 'Use when' is left unfinished and does not clearly define invocation boundaries, prerequisites, or consent requirements. Ambiguous activation increases the chance that the agent applies this skill in the wrong context and executes setup or speech-generation steps without sufficient user intent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
71% confidence
Finding

The description states "支持中文交互" while the rest of the file mixes English and Chinese instructions, but it does not clearly offer the user a language or locale choice. This can create an implicit language behavior that is not explicitly opt-in.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill claims 'zero cloud calls' and 'works completely offline after setup', but its documented setup explicitly downloads a default voice and later points users to Hugging Face for more voices. This mismatch can mislead operators in restricted or no-network environments, causing unexpected external connections, policy violations, or supply-chain exposure during installation.

Content

No source excerpt is available for this finding.