The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts located in the user's dotfiles directory to manage session state.
Evidence: Found calls to '~~/code/seabbs/dotfiles/scripts/claude-conductor-gather.sh' and '~~/code/seabbs/dotfiles/scripts/claude-session-track.sh'.
Context: The skill includes logic to verify the script's existence and permissions before execution.
[DATA_EXFILTRATION]: The skill reads and displays the session history (prompts and assistant responses) from other active Claude Code sessions.
Evidence: The gathering script returns 'recent_prompts' and 'last_response' fields which are displayed in the dashboard and drill-down views.
Context: This cross-session data access is the primary purpose of the skill and is confined to the local environment.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by ingesting untrusted data from session histories into the current agent context.
Ingestion points: 'recent_prompts' and 'last_response' fields in JSON output from the gathering script.
Boundary markers: Absent; data is rendered directly into Markdown tables and text blocks.
Capability inventory: Execution of local scripts and writing to the file system at '~/.claude/inbox/'.
Sanitization: No explicit sanitization or filtering is defined for the ingested session data.

conductor