The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes session history, including user inputs and command outputs, to generate summaries that drive tool execution and new agent tasks, creating a surface for indirect prompt injection. * Ingestion points: The skill ingests the entire session history (conversation, command outputs, and file content) into a summary as described in SKILL.md (Workflow Step 2) and references/subagent-spawn-pattern.md. * Boundary markers: No specific delimiters or safety instructions are used to isolate the ingested session data from the agent's instructions. * Capability inventory: The skill utilizes the context-file tool for filesystem writes and supports spawning exploration sub-agents with dynamically generated prompts. * Sanitization: No sanitization or escaping of the session content is performed before its use in summaries or document creation.
[DYNAMIC_EXECUTION]: The skill implements a pattern for dynamically spawning sub-agents with prompts constructed at runtime based on the current session state. * Evidence: references/subagent-spawn-pattern.md outlines the workflow for spawning an exploration sub-agent using a template that incorporates a dynamically generated session summary.

session-reflection