The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages knowledge vaults by executing a local Python script (claude-obsidian.py) for setup, migration, and data persistence tasks.
Evidence: Commands like python3 "$CORE" init and python3 "$CORE" transaction apply are documented in SKILL.md and references/operation-transactions.md.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes content from external sources such as web research and browser clippings.
Ingestion points: Web Clipper functionality described in references/plugins.md and the autoresearch route mentioned in SKILL.md.
Boundary markers: The use of structured frontmatter (references/frontmatter.md) and SHA-256 hashes in transaction bundles (references/operation-transactions.md) serves to delimit and verify ingested data.
Capability inventory: The skill can execute local commands via the Python core and write files within the vault's directory structure.
Sanitization: The skill's documentation explicitly requires path validation, integrity checks, and advises treating all external clippings as untrusted input to be reviewed before ingestion (references/rest-api.md, references/plugins.md).

wiki