ai-adr
Pass
Audited by Gen Agent Trust Hub on Nov 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool for repository maintenance and file operations, including directory creation (
mkdir), file relocation (mv,git mv), and path searching (find). These operations are strictly limited to the local project environment via theallowed-toolsconfiguration, preventing arbitrary command execution. - [DATA_EXFILTRATION]: The skill accesses local Git configuration data (
user.name,user.email) to attribute authorship within the Architecture Decision Records. This information is used exclusively to populate local markdown files and is not transmitted to external services or endpoints. - [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied decision descriptions and existing ADR files to maintain the project log. It mitigates risks by requiring explicit human invocation (
/ai-adr), providing structured templates, and using clear markers (— *(inferred)*) for content not directly supplied by the user. The skill also explicitly forbids the agent from invoking the log management commands autonomously.
Audit Metadata