update-patterns
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from user notes to synthesize patterns, which presents an indirect prompt injection surface.\n
- Ingestion points: The agent is instructed to read metadata and body content from markdown notes as specified in the execution protocol in SKILL.md.\n
- Boundary markers: There are no explicit markers or instructions to the agent to treat note content as untrusted data during the analysis and synthesis phase.\n
- Capability inventory: The update_patterns.py script has the capability to read, write, modify, and append content to markdown files within the user's Obsidian vault.\n
- Sanitization: While the script validates the structure of the JSON payload and slugifies IDs, it does not sanitize or filter the natural language content provided by the agent before writing it to the filesystem.\n- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute specific local Python scripts to perform updates and validation.\n
- The agent is instructed to run update_patterns.py to apply updates to pattern files via uv run python.\n
- The agent is instructed to run generate_metadata.py to validate incoming notes before processing.\n
- These scripts are located in paths corresponding to the author's local environment.
Audit Metadata