conductor
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes a local shell script located at
~/code/seabbs/dotfiles/scripts/claude-conductor-gather.shto retrieve active session data. While the script is part of the user's local dotfiles (associated with the skill author), executing scripts at runtime is a privileged operation. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and displays historical data from other Claude sessions, creating a vector where malicious content in those sessions could influence the current agent.
- Ingestion points: The
recent_promptsandlast_responsefields are read from the output of the gathering script and displayed in the dashboard and drill-down views inSKILL.md. - Boundary markers: The ingested data is placed within a Markdown table and text list without specific instructions or delimiters to treat the content as untrusted data.
- Capability inventory: The skill has the capability to write files to the filesystem (
~/.claude/inbox/{session-id}.md) using the Write tool. - Sanitization: There is no mention of sanitizing, escaping, or filtering the session history data before it is presented to the agent.
Audit Metadata