The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates shell commands to interact with the Obsidian CLI by interpolating content derived from the conversation into command arguments. In Step 5, the obsidian create and obsidian append commands use a content parameter formatted as content="<full note content>". Because this content is generated from conversation history, any shell metacharacters such as double quotes, backticks, or dollar signs included in the text could break out of the string literal and lead to arbitrary command execution on the host machine.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes untrusted conversation history to drive file-system and shell-level operations.
Ingestion points: Conversation history extracted in Step 3 and user-supplied topic names from $ARGUMENTS in Step 1.
Boundary markers: None. The skill instructions do not specify any delimiters or safety prompts to treat the extracted content as untrusted data.
Capability inventory: The skill possesses the ability to execute shell commands (pwd, git remote, obsidian) and write to the local file system.
Sanitization: None. There is no instruction to escape or validate the content before it is interpolated into shell command strings.

til