afk
Pass
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes multiple shell scripts from a local directory (e.g.,
bin/fm-afk-start.sh,bin/fm-supervise-daemon.sh) and uses system utilities likedate. It also interacts directly with terminal environments usingsend-keysfor tmux orpane send-textfor herdr to automate input.\n- [PROMPT_INJECTION]: The skill establishes an indirect prompt injection vector by design. It reads status information and event logs from external files (state/.wake-queue,state/.subsuper-escalations) and injects these summaries into the agent's prompt context.\n - Ingestion points: External data is ingested from files in the
state/directory.\n - Boundary markers: Uses a specific ASCII control character (0x1f) as a sentinel marker to distinguish injected content from user messages.\n
- Capability inventory: The skill can execute local scripts and send text to the terminal harness.\n
- Sanitization: Implements basic sanitization by collapsing newlines in the injected text to maintain a single-line format.
Audit Metadata