The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is specifically designed to access and retrieve historical conversation data, which includes user prompts, assistant responses, tool usage logs, and file interaction history. While this is the primary purpose of the skill, the data retrieved is highly sensitive.
[PROMPT_INJECTION]: The skill introduces a surface for indirect prompt injection by retrieving and incorporating historical session data (transcripts and AI-generated outlines) into the current conversation context. Malicious instructions or adversarial content present in a recalled session could influence the agent's behavior in the current session.
Ingestion points: Historical session data retrieved via scripts/transcript and scripts/search (which provides AI-generated outlines).
Boundary markers: The transcript format uses structural tags like [U] (User), [A] (Assistant), and [T] (Tool) to delimit different message types, but lacks strong isolation for embedded instructions.
Capability inventory: The agent utilizing this skill typically has access to powerful tools such as Edit (file system modification) and Bash (command execution), as evidenced by the transcript examples.
Sanitization: The scripts retrieve raw or formatted text from the backend server without evidence of content sanitization or filtering to prevent instruction re-activation.

session-recall