The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides endpoints to read and update high-leverage settings such as system prompts and payload mutation rules, which constitutes an attack surface for indirect prompt injection.
Ingestion points: GET /api/settings/system-prompt and GET /api/settings/payload-rules as described in SKILL.md and references/endpoints.md.
Boundary markers: The skill does not provide specific instructions or delimiters for the agent to use when processing the content of these settings.
Capability inventory: The skill enables the agent to perform sensitive updates to system prompts and mutation rules via PUT requests, which can fundamentally alter the managed application's behavior.
Sanitization: While the API is documented to perform regex validation and error sanitization, no specific prompt-level sanitization for system prompt content is mentioned.
[SAFE]: The skill consists entirely of markdown documentation and endpoint references for a local management API. It does not include executable code, remote script downloads, or obfuscated instructions.

omni-settings