show-me-your-work
Pass
Audited by Gen Agent Trust Hub on Aug 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a local utility script,
scripts/log.sh, to append decision entries to a TSV file. The script performs standard filesystem operations such as directory creation and file writing without requiring elevated privileges. It avoids dangerous patterns likeevalor unsanitized command interpolation. - [DATA_EXPOSURE]: The instructions mandate an audit process where the agent reads its own session transcripts from the
agent-transcripts/directory. The skill includes specific safeguards to prevent the agent from over-reaching into unrelated project directories or private user data. - [INDIRECT_PROMPT_INJECTION]: The
scripts/log.shscript includes a dedicatedclean()function that mitigates CSV/Formula Injection risks. It identifies inputs starting with formula-triggering characters (=,+,-,@) and escapes them with a single quote, ensuring that malicious content in logged evidence (e.g., an attacker-controlled PR title) cannot execute commands when the TSV log is opened in spreadsheet applications.
Audit Metadata