handoff-execute
Warn
Audited by Gen Agent Trust Hub on Jul 16, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes arbitrary shell commands specified in the
@OUTCOME.verifyfield of the input handoff brief during Phase 5. - [COMMAND_EXECUTION]: It automatically runs project-defined gate suites, such as
turbo run lint test typecheck, which involves executing local binary or script paths. - [EXTERNAL_DOWNLOADS]: The skill fetches external content from URLs listed in the
@SOURCESblock of the handoff brief using the runtime's web tool. - [REMOTE_CODE_EXECUTION]: By interpreting and executing commands from an external brief file, the skill performs dynamic code execution based on external data inputs.
- [DATA_EXFILTRATION]: The skill reads arbitrary files requested by the brief's
@SOURCESsection and records their metadata, hashes, and git diffs into a 'retro' file written to the local vault. - [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from the handoff brief (Ingestion point:
<brief-path>) using YON block delimiters (@TASK,@OUTCOME) as boundaries. It possesses significant capabilities including subprocess execution and file system access, but lacks explicit sanitization of the brief's content.
Audit Metadata