The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is specifically designed to process untrusted external data from sources such as email imports, calendar invites, and chat transcripts to update an internal knowledge base. It lacks instructions for sanitizing these inputs or using boundary markers to prevent instruction injection.
Ingestion points: Processes content from email archives, calendar imports, chat transcripts, and meeting notes as described in the "No Placeholders at Ingest" section.
Boundary markers: The instructions do not specify the use of delimiters or warning prompts to treat the external data strictly as non-executable text.
Capability inventory: The skill utilizes gbrain think for synthesis, gbrain search for lookups, and gbrain put_page for writing resolved data back to the file system (people/, companies/).
Sanitization: There is no evidence of data sanitization before interpolation into command arguments or during the write-back process, allowing for potential data poisoning.
[COMMAND_EXECUTION]: The skill utilizes a suite of local CLI tools (gbrain) to perform operations. The commands interpolate variables (such as {entity} and {entity-slug}) that are derived from untrusted external sources, which may lead to command injection if the underlying shell or CLI tool does not properly escape these values.

resolve-before-asking