The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data that is later stored and potentially processed by other agent skills.
Ingestion points: Untrusted data enters the skill via the $ARGUMENTS variable in SKILL.md, which contains the raw text body provided by the user.
Boundary markers: There are no boundary markers or delimiters used when saving the content to the vault to prevent downstream agents from interpreting the note body as instructions.
Capability inventory: The skill utilizes the Bash Write tool to execute new_note.sh, which performs file system write operations within the $OBSIDIAN_VAULT directory.
Sanitization: While the code sanitizes the SLUG used for the filename, the --title and the stdin BODY are passed to the writing script without escaping or filtering malicious instructions.
[COMMAND_EXECUTION]: The skill executes a local shell script located at a relative path (../obsidian-vault/scripts/new_note.sh). This script is used to enforce the vault's schema and structure during note creation.

ingest-quick-note