The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to manage the environment and perform API requests.
Evidence:
Execution of chmod 600 on the API key file ~/.pixazo/api-key to restrict access.
Use of curl to perform POST requests to the Pixazo API gateway.
[EXTERNAL_DOWNLOADS]: The skill fetches content from external sources for configuration and reference.
Evidence:
Downloads additional instructions and schema references from https://www.pixazo.ai/models/seedvr.md (a vendor-owned resource).
The README suggests using npx skills add Pixazo-AI/skills to install related tools from the same author.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to instructions embedded in external data it is instructed to process.
Evidence Chain:
Ingestion points: Fetches instruction updates and parameter schemas from the external URL https://www.pixazo.ai/models/seedvr.md at runtime (noted in SKILL.md).
Boundary markers: Absent. There are no instructions for the agent to ignore potentially malicious directions within the fetched markdown file.
Capability inventory: The agent has capabilities for network operations (curl, requests, fetch) and local file system writes (~/.pixazo/api-key).
Sanitization: Absent. The skill does not define validation for the content retrieved from the remote documentation URL.

seedvr