The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill creates a handoff artifact that serves as an instruction set for subsequent agent sessions. It explicitly includes a 'Handoff Prompt for Next Agent' that directs the next agent to 'Preserve listed decisions and constraints exactly as written' and 'execute "Next Steps (Ordered)" in sequence'.
Ingestion points: The skill ingests the entire conversation history, including user messages, assistant responses, and command outputs, to populate the handoff template.
Boundary markers: While the template uses Markdown headers and source tags, it lacks explicit sanitization or instructions for the agent to ignore potential commands embedded within the summarized data.
Capability inventory: The skill uses file-writing capabilities to save the snapshot to the local filesystem (e.g., handoffs/context-YYYY-MM-DD-HHMM.md).
Sanitization: No sanitization, escaping, or filtering of external content is performed before it is interpolated into the instructions for the next agent.
[DATA_EXPOSURE]: The skill's primary function is to record high-signal context, which includes command results and file identifiers. If a session involves sensitive data—such as API keys in command output or secrets in referenced files—this data will be persisted in the handoff file within the handoffs/ directory, increasing the risk of accidental credential exposure.

context-snapshot