The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill workflow involves reading recent session notes and workspace files to generate summaries. This creates a surface where the agent could ingest data containing malicious instructions. However, the risk is minimal because the skill does not grant the agent access to dangerous tools (such as network operations or arbitrary command execution) that could be used for exploitation.
[DATA_EXPOSURE]: The skill includes a 'Quality Bar' section that explicitly instructs the agent to be 'privacy-safe' by excluding secrets, credentials, and private personal details from the handoff notes, which is a positive security practice.

agent-handoff