The Agent Skills Directory

[PROMPT_INJECTION]: The skill has a potential surface for indirect prompt injection as it processes content from external files that may be modified by other users or processes.
Ingestion points: The agent reads task titles, descriptions, and metadata from the .simbl/tasks.md file using the simbl list and simbl show commands.
Boundary markers: There are no explicit markers or instructions provided to the agent to treat task content as untrusted data or to ignore embedded instructions.
Capability inventory: the agent uses the Bash tool to interact with the file system and execute CLI commands.
Sanitization: No sanitization or validation of the content read from the task files is specified in the documentation.
[EXTERNAL_DOWNLOADS]: The skill instructions involve downloading and installing an external software package.
Evidence: The guide includes commands to globally install the simbl package via npm install -g simbl or bun install -g simbl.
Context: The package is installed without a specified version and is fetched from the public npm registry.

cli-guide