short
Warn
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill invokes
./scripts/skillsmp-search.sh "<task type>" --installto search for and install specialized tools or scripts at runtime based on the current task. This pattern allows for the dynamic acquisition and execution of unverified code. - [COMMAND_EXECUTION]: The skill performs multiple shell operations including reading git history (
git log --oneline -5), verifying tool availability (command -v codex), and running user-defined tests and linters. It also interacts with the Codex CLI via bash. - [EXTERNAL_DOWNLOADS]: The skill uses the
/doc <name> <url>pattern to fetch documentation or content from external URLs to populate a local cache. This allows for the ingestion of data from arbitrary remote sources. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by reading and following instructions found in local project files such as
1shot/DECISIONS.md,1shot/BLOCKERS.md, and git commit messages. These sources are not sanitized and could influence agent behavior if they contain malicious instructions.
Audit Metadata