The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the jj (Jujutsu) CLI tool to retrieve code changes and file structures. Specifically, it executes jj diff -s and jj diff -- [path] to gather context for the review process.
[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection, as it reads external source code and incorporates it directly into the agent's reasoning context.
Ingestion points: Untrusted data is ingested via jj diff commands in SKILL.md and SUBAGENT.md.
Boundary markers: Absent. The skill does not wrap the ingested code in delimiters or provide instructions to the agent to ignore any commands embedded within the code.
Capability inventory: The agent can execute CLI commands (jj), access the web via WebSearch/WebFetch, and orchestrate sub-tasks using a Task tool.
Sanitization: No sanitization or filtering is applied to the source code before it is processed by the main agent or the sub-agents.

code-review