The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation recommends installing odino, a third-party CLI tool, via pipx. This package does not originate from a trusted or well-known organization as defined in the security policy.
[COMMAND_EXECUTION]: Bash command templates provided in SKILL.md and the reference guides (e.g., odino query -q "$QUERY") directly interpolate variables derived from user input into shell commands. This pattern is susceptible to command injection if a malicious user provides a query containing shell metacharacters.
[PROMPT_INJECTION]: The skill facilitates indirect prompt injection by ingesting untrusted data from the user's codebase during search operations.
Ingestion points: odino query output tables and subsequent read or cat operations on files discovered via semantic search.
Boundary markers: Absent. The skill does not implement delimiters or instructions to ignore potential commands embedded within the code or documentation it searches.
Capability inventory: The skill utilizes Bash (allowing arbitrary command execution) and Read tools.
Sanitization: Absent. No logic is provided to escape or validate the contents of the files before they are presented to or processed by the agent.
[DATA_EXFILTRATION]: The skill's design explicitly encourages searching for sensitive information, with search patterns specifically targeting "API key validation" and "password hashing and verification." This increases the risk of sensitive credentials being exposed in the agent's context or conversation history.

semantic-search