create-spec
Pass
Audited by Gen Agent Trust Hub on Apr 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the way it processes external research files.\n
- Ingestion points: Data is read from the
research/directory or a location specified by the user in$ARGUMENTS(referenced inSKILL.md).\n - Boundary markers: Absent. There are no delimiters or instructions to prevent the agent from obeying commands or instructions embedded within the research documents.\n
- Capability inventory: The skill can write files to the
specs/directory and use theAskUserQuestiontool to prompt the user (as defined inSKILL.md).\n - Sanitization: Absent. The skill does not validate or sanitize the content of the research documents before incorporating them into the final specification.\n- [COMMAND_EXECUTION]: The skill utilizes dynamic context injection to execute a local shell command at load time.\n
- Evidence:
!git config user.name`` inSKILL.mdused to populate the author field in document templates.\n - Context: This command retrieves the local git user configuration for document personalization, which is a common and benign developer workflow.
Audit Metadata