The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the external obsidian-wiki CLI tool to perform data processing, clustering, and updating names within the local session history database. It also uses the open command to launch a web browser to display the generated graph.
[DATA_EXFILTRATION]: The skill accesses sensitive user data stored in the ~/.claude/ directory, specifically reading all session transcripts and the history.jsonl file to generate its topic graph.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from past session transcripts to perform summarization and clustering tasks.
Ingestion points: Reads transcripts and history.jsonl from the local ~/.claude/ directory.
Boundary markers: None identified; instructions do not provide delimiters or warnings to ignore instructions embedded within the history files.
Capability inventory: Executes shell commands via obsidian-wiki and launches a browser with open.
Sanitization: No evidence of sanitization, filtering, or escaping of the transcript content is present before the data is processed by the agent to name clusters.

session-brain