The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from a knowledge base vault (e.g., downloaded papers or external notes). Malicious instructions within these documents could attempt to hijack the agent's synthesis process or influence the creation of new documents during the 'Accumulate' phase.
Ingestion points: Accesses and reads files across various PARA folders (Projects, Areas, Resources, Archives) as described in the Execution Flow and Route Selection sections.
Boundary markers: The skill does not define explicit delimiters or instructions to the agent to disregard instructions found within the source documents during synthesis.
Capability inventory: Performs file reads and searches via grep and the obsidian CLI, and has the capability to write new notes or update index files such as _index.md and log.md (documented in the Accumulate and Execution Flow steps).
Sanitization: The instructions lack specific mechanisms to sanitize or escape data retrieved from the vault before presenting it to the user or incorporating it into new system files.

kb-query