The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes conversation history to create summaries, creating a surface for indirect prompt injection if the history contains malicious instructions.
Ingestion points: The skill utilizes the current conversation history as its primary input in SKILL.md.
Boundary markers: There are no explicit delimiters or boundary instructions provided to separate conversation data from the summary logic.
Capability inventory: The skill has the capability to write files to the local filesystem.
Sanitization: The instructions explicitly require the agent to redact sensitive information, such as API keys and passwords, before saving.
[COMMAND_EXECUTION]: The skill directs the agent to execute a synchronization command upon completion.
Evidence: The instruction to run /vault-sync in the Vault section of SKILL.md.
[DATA_EXFILTRATION]: The skill instructs the agent to save summarized conversation data to a location outside the project workspace.
Evidence: Instruction to `Save to the temporary directory of the user's OS
not the current workspaceinSKILL.md`. While the skill includes redaction steps, saving data to system-wide temporary directories can lead to unintended exposure in shared environments.

handoff