The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a custom, non-standard binary located at a hardcoded user path: /Users/mtm/go/bin/islandiguana. Executing arbitrary binaries from user-controlled directories is a risk as the binary's integrity and behavior are not verified.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to perform in-place mutations (update, append, delete) of Markdown files using yq expressions, which allows for the modification or potential corruption of local data.
[DATA_EXFILTRATION]: The skill is designed to recursively read and process the contents of a user's Obsidian vault located at /Users/mtm/Documents/Obsidian Vault, which may contain sensitive personal information or private documents.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface because it ingests data from local Markdown files and provides them to the agent's context without sanitization.
Ingestion points: Markdown files in the Obsidian vault as specified in the SKILL.md instructions.
Boundary markers: No boundary markers or 'ignore embedded instructions' warnings are present to distinguish between vault content and agent instructions.
Capability inventory: File mutation (edit/delete) via the islandiguana tool, shell command execution via Bash (including rg, grep, and xargs), and the ability to read arbitrary file contents.
Sanitization: No sanitization, filtering, or validation is performed on the data retrieved from the vault before it is processed by the agent.

islandiguana