The Agent Skills Directory

[SAFE]: The skill implements a structured documentation and memory management system without any malicious patterns. It outlines clear logic for knowledge consolidation and includes security-conscious examples, such as utilizing hooks to prevent writes to sensitive files like .env or credential stores.
[SAFE]: Although the skill processes data from external sources (an Obsidian vault and local rule files), which represents a surface for indirect prompt injection, the behavior is within the expected scope for memory management and targeting local or user-curated files.
Ingestion points: The skill reads from memory/rules/cc-always.md, memory/rules/cc-patterns.md, and Obsidian vault paths such as Research/Claude-Code/Patterns/*.md and Repositories/{org}/{repo}.md.
Boundary markers: No explicit delimiters or boundary markers to isolate ingested content from agent instructions were identified.
Capability inventory: The skill utilizes memory management tools (mem_save, mem_search, mem_get_observation, mem_context), Obsidian MCP tools (mcp__obsidian__*), and direct file system write capabilities.
Sanitization: No evidence of explicit sanitization, validation, or escaping of the ingested content was found.

cc-second-brain