impact-analysis
Pass
Audited by Gen Agent Trust Hub on Mar 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill extensively uses the Bash tool to run local search operations using
grepandfind. It also performs file system management, such as creating theproject-decisions/directory and generating an index file usingmkdir,echo, and redirection. - [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface (Category 8). It ingests content from the local codebase (source code, comments, and documentation) using
grepandhead. This data is then used to populate the 'Impact Analysis' reports. - Ingestion points: Codebase files processed via
grepandhead(SKILL.md). - Boundary markers: Absent; the content read from the files is directly used in the report structure without delimiters.
- Capability inventory: File system write access via
Bash(mkdir, echo) and read access via search tools. - Sanitization: Absent; the skill uses
sedfor output formatting but does not sanitize content to prevent instruction injection from the analyzed files.
Audit Metadata