idea-capture
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions direct the agent to read and prioritize conventions defined in the vault's
CLAUDE.mdfile. This creates a surface for indirect prompt injection, where an attacker could place malicious instructions inCLAUDE.mdto manipulate the agent's behavior during the idea capture process.\n - Ingestion points: The agent reads content from
CLAUDE.md,Wiki/index.md, and existing notes inIdeas/andWiki/concepts/to determine status, duplicates, and related concepts.\n - Boundary markers: No boundary markers or specific safety prompts are utilized to isolate data ingested from these vault files from the agent's instruction context.\n
- Capability inventory: The skill allows the agent to create new Markdown files in
Ideas/and append entries toInbox/Tasks.md. These actions are performed by a local Python script (scripts/idea.py).\n - Sanitization: The Python script implements proper filename slugification and YAML string escaping, providing robust protection against path traversal and metadata injection.
Audit Metadata