secondmate-provisioning

Pass

Audited by Gen Agent Trust Hub on Jul 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on executing a suite of local shell scripts (e.g., bin/fm-brief.sh, bin/fm-home-seed.sh, bin/fm-spawn.sh, bin/fm-teardown.sh) to perform system administration, process management via tmux, and file system worktree manipulation.
  • [PROMPT_INJECTION]: The skill manages agent behavior by processing task data from data/backlog.md and data/secondmates.md, creating a surface for indirect prompt injection. 1. Ingestion points: Processes external task data from data/backlog.md and configuration metadata from data/secondmates.md. 2. Boundary markers: No explicit delimiters or instructions to ignore embedded commands are defined for the processed backlog content. 3. Capability inventory: Executes subprocesses for agent spawning and teardown; performs file-system operations including writing and copying configuration files. 4. Sanitization: The instructions do not specify sanitization or escaping when interpolating natural language strings for the charter or scope into shell environment variables (e.g., FM_SECONDMATE_CHARTER='<charter>').
  • [EXTERNAL_DOWNLOADS]: The skill utilizes a treehouse command to manage durable worktree leases.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 16, 2026, 06:14 PM