experiment-craft
Pass
Audited by Gen Agent Trust Hub on Mar 17, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection by processing external data as part of its diagnostic flow.\n
- Ingestion points: The skill uses
read_fileto collect failure cases and analyze results from experiment logs as described inSKILL.mdandreferences/debugging-methodology.md.\n - Boundary markers: The instructions do not specify any delimiters or warnings to ignore commands embedded in the processed data.\n
- Capability inventory: The skill has access to
write_file,edit_file, andexecutetools (allowed inSKILL.md) to perform actions based on the analysis.\n - Sanitization: There is no evidence of sanitization or validation of the ingested experiment data.\n- [COMMAND_EXECUTION]: The skill uses the
executetool to run code for verifying experimental hypotheses and implementation fixes, which is a high-privilege capability necessary for its function.
Audit Metadata