The Agent Skills Directory

[SAFE]: The skill is composed entirely of natural language instructions designed to assist in summarizing session context. It does not perform any sensitive operations.
[DATA_EXFILTRATION]: No network-capable tools or commands (curl, wget, fetch) are used. The skill explicitly states it does not write to disk; its output is intended for the user to copy or save manually.
[COMMAND_EXECUTION]: While the skill mentions reviewing git diff, this is a common read-only operation for tracking changes within a repository and is used here solely for information gathering.
[PROMPT_INJECTION]: No patterns associated with system prompt extraction, safety filter bypass, or instruction overrides were detected in the instructions or metadata.
[NO_CODE]: The skill contains no executable scripts (Python, Node.js, Shell, etc.) and relies entirely on the underlying agent's reasoning capabilities.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or credentials were found in the skill content.
[EXTERNAL_DOWNLOADS]: The skill does not reference or download any external resources or dependencies.

session-handoff