The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from the Dinox CLI and note database while having the capability to perform hierarchy modifications, creating a potential surface for indirect prompt injection.
Ingestion points: The agent reads CLI output from dino tag list, dino tag tree, dino tag stats, and dino note search as per instructions in the 'Intent Routing' section of SKILL.md.
Boundary markers: The instructions mandate the use of --format json for structured output and explicitly require the agent to treat all data from Dinox (notes, tags, CLI output) as untrusted data.
Capability inventory: The skill utilizes powerful CLI commands including dino tag add, dino tag rename, dino tag move, dino tag merge, and dino tag cleanup as listed in the 'Command Reference'.
Sanitization: The runtime contract requires that user-provided values are kept as individual arguments to avoid shell interpolation, and it recommends using secure temporary files for processing free-form note content.
[COMMAND_EXECUTION]: The skill is designed to interact with the host system by executing the dino CLI tool to manage application state.
Details: The skill provides a comprehensive command reference for the dino binary. This execution is the primary intended function of the skill and is governed by strict safety rules, such as enforcing dry-runs and requiring user confirmation before performing merges or deletions.
[DYNAMIC_CONTEXT_INJECTION]: The instructions reference the platform-specific !command syntax for authentication.
Evidence: SKILL.md documents that users in specific environments like Claude Code can run ! dino auth login to authenticate.
Analysis: This mention is instructional and documents the official authentication flow for the dino tool. It does not represent an attempt to execute hidden or malicious commands at skill load time.

dino-manage-tags