The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill provides bash command templates that perform file movement and permanent deletion.
Evidence: The file SKILL.md contains commands like rm -rf "/path/to/Obsidian Vault/SourceFolder" and cp -r "/path/to/Obsidian Vault/SourceFolder/"*.
Risk: If the agent incorrectly identifies the vault path or is manipulated into targeting sensitive system directories, it could cause irreversible data loss.
[INDIRECT_PROMPT_INJECTION] (MEDIUM): The skill is designed to process external content (local filesystem directory structures) and possesses the capability to modify or delete files.
Ingestion points: SKILL.md uses ls -la and find to ingest directory structures.
Boundary markers: None identified. The skill does not provide instructions to distinguish between metadata and potential malicious content within folder names.
Capability inventory: mkdir, cp, rm -rf.
Sanitization: The skill uses double-quoting in shell examples, but the security depends entirely on the underlying agent's ability to safely interpolate user-provided paths into these commands without introducing shell injection vulnerabilities.

obsidian-organize