audio-slice-api
Warn
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: MEDIUMMETADATA_POISONINGCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The skill's description and Step 4 instructions claim it is for 'Image generation/editing,' yet the skill name ('audio-slice-api'), triggers, and primary API implementation ('audio-slice') are for audio trimming. This contradiction is deceptive and can cause the agent to be triggered incorrectly or provide confusing results to the user.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands, including 'chmod 600' for file permission management and 'npx skills add Pixazo-AI/skills' for environment setup.
- [EXTERNAL_DOWNLOADS]: The skill provides an 'npx' command that downloads and executes code from the 'Pixazo-AI/skills' repository. This introduces a dependency on external code that is fetched at runtime.
- [REMOTE_CODE_EXECUTION]: The instruction to run 'npx skills add Pixazo-AI/skills --skill "*"' constitutes remote code execution, as it pulls scripts from a remote source and executes them in the user's environment.
- [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the ingestion of external data.
- Ingestion points: API responses from 'gateway.pixazo.ai' (specifically JSON error bodies) and external audio URLs provided by the user.
- Boundary markers: Absent. The agent is instructed to surface API error messages 'verbatim' and to render resulting URLs directly to the user.
- Capability inventory: The skill possesses file-write capabilities ('~/.pixazo/api-key'), shell execution capabilities ('curl', 'npx'), and network access ('gateway.pixazo.ai').
- Sanitization: There is no evidence of sanitization or validation performed on the contents of API responses before they are displayed or processed.
Audit Metadata