The Agent Skills Directory

[METADATA_POISONING]: The skill exhibits significant internal contradictions between its stated purpose and its implementation details. While titled 'audio-normalize-api' and targeting a 'media-normalize-audio' endpoint, the description and instructions repeatedly refer to image generation and editing tasks. Specifically, the 'Trigger' section and 'Step 4' contain instructions for handling images, 'image_size' enums, and 'seed' parameters, which are inconsistent with the audio normalization functionality described elsewhere.
[EXTERNAL_DOWNLOADS]: The skill provides a command in the 'Related Pixazo skills' section (npx skills add Pixazo-AI/skills) that fetches and executes code from a remote repository. While the repository belongs to the skill's authoring organization (Pixazo-AI), it represents a remote code execution pattern.
[COMMAND_EXECUTION]: The agent is instructed to execute shell commands, including curl for making API requests, chmod 600 for file permission modification, and npx for skill installation.
[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by processing untrusted data from external sources and user inputs.
Ingestion points: The skill ingests data from user-provided URLs (audio_url) and prompts within the SKILL.md workflows.
Boundary markers: There are no clear delimiters or instructions for the agent to ignore potentially malicious content embedded in the processed audio URLs or prompts.
Capability inventory: The skill has the ability to write to the file system (~/.pixazo/api-key), execute shell commands (curl, npx, chmod), and perform network operations (requests.post, fetch).
Sanitization: No input sanitization or validation logic is defined to check the integrity of the external data before it is interpolated into commands or prompts.

audio-normalize-api