The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for environment discovery and file management. It uses pgrep and lsof to inspect running processes and identify those to be handed off or terminated. It uses a shell script block to generate a project hash, create directories, and manage files (mkdir, cp, cat). It also uses pbcopy to move distilled session data into the system clipboard, which can be accessed by other applications.
[DATA_EXFILTRATION]: The skill saves session summaries to a persistent, hidden location in the user's home directory (~/.claude/claude-code-token-saver-data/). While intended for context restoration, this behavior stores session-specific findings outside the project's own directory structure. The use of pbcopy also exposes data to the system clipboard.
[PROMPT_INJECTION]: The skill aggregates unverified data from subagents and tools into a handoff file designed to be automatically loaded in future sessions, creating a surface for indirect prompt injection.
Ingestion points: The skill (SKILL.md) explicitly walks the current session to extract data from subagent findings and tool outputs.
Boundary markers: The skill does not define specific delimiters or instructions to ignore embedded commands within the extracted data.
Capability inventory: The skill possesses capabilities for shell command execution, file writing, and clipboard manipulation.
Sanitization: There is no evidence of sanitization or filtering of the extracted content before it is written to the handoff file.

cc-compact