The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to gather evidence from the current task and session to populate summaries, decisions, and follow-ups. If the data being summarized contains malicious instructions (e.g., from a file the agent previously analyzed), those instructions could be included in the generated handoff document.
Ingestion points: SKILL.md workflow step 1 requires using "current task's evidence" to record completed work.
Boundary markers: There are no instructions for the agent to use delimiters or to disregard instructions found within the session evidence.
Capability inventory: The scripts/session-done script can write to the local filesystem via the --output argument.
Sanitization: The Python script scripts/session-done performs no sanitization or validation of the text content before rendering it into the Markdown output.
[COMMAND_EXECUTION]: The skill uses a local Python script scripts/session-done to format and optionally persist session data. The script uses standard libraries and implements safety checks to avoid overwriting existing files.

session-done