The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting data from potentially untrusted local project files and propagating it to other systems.
Ingestion points: The skill reads metadata and session summaries from CLAUDE.md and log files (e.g., log/YYYY-MM-DD-HHMM.md) located within the project directory.
Boundary markers: There are no specific delimiters or instructions for the agent to ignore embedded commands within the content it extracts from these project files.
Capability inventory: The skill has the capability to write to local files (Edit, Write tools) such as .context/projects/_index.md and .context/current-focus.md, and to modify remote content via Notion MCP tools (mcp__claude_ai_Notion__notion-update-page).
Sanitization: No sanitization, escaping, or validation logic is defined to check the extracted content before it is used to update local registries or the Notion Research Pipeline.

sync-notion