agentic-jujutsu
Pass
Audited by Gen Agent Trust Hub on May 10, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill utilizes the
agentic-jujutsupackage from the NPM registry and references the@qudag/napi-coredependency for cryptographic operations. - [COMMAND_EXECUTION]: Provides functionality via the
jj.execute()method to run shell commands, primarily intended for version control operations such as git pushes and merges. - [PROMPT_INJECTION]: The skill features a learning system where user-provided task descriptions are processed by
getSuggestionto recommend operations. This represents an indirect prompt injection surface where untrusted input could influence agent behavior. - Ingestion points: Task descriptions passed to
getSuggestion(task)andqueryTrajectories(task, limit)as seen inSKILL.md. - Boundary markers: No specific boundary markers or instruction isolation mechanisms are documented in the provided implementation examples.
- Capability inventory: Automated execution of shell commands through the
jj.execute()method. - Sanitization: No input validation or sanitization routines are specified for the ingested task descriptions.
Audit Metadata