weekly-synthesis

Pass

Audited by Gen Agent Trust Hub on Jun 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill incorporates defensive design patterns such as explicit user confirmation before executing write operations (obsidian_write_note) and verification of results via read-back operations.
  • [COMMAND_EXECUTION]: The skill includes local shell scripts (validators/pre-check.sh and validators/post-check.sh) used for validating the environment and output structure. These scripts perform benign text analysis and environment checks with no network activity or sensitive file access.
  • [PROMPT_INJECTION]: The skill processes user-controlled data from Obsidian notes. It mitigates indirect prompt injection risks by using structured templates (found in references/templates.md) and a phased analysis approach that separates data ingestion from synthesis logic.
  • Ingestion points: Vault notes are read via obsidian_get_note in SKILL.md.
  • Boundary markers: The procedure uses distinct phases and reporting headers to isolate note content.
  • Capability inventory: File write access is limited to obsidian_write_note and requires user confirmation.
  • Sanitization: Output is constrained to specific fields (One action, One question) and validated by post-execution scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 29, 2026, 06:30 AM
Security Audit — agent-trust-hub — weekly-synthesis