The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the jj (Jujutsu) command-line interface to determine which files have changed in the repository before running reviews.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted code files and passes them to LLM-based agents.
Ingestion points: Files identified by the jj diff command in both SKILL.md and SUBAGENT.md.
Boundary markers: Absent; the instructions do not use delimiters or provide "ignore embedded instructions" warnings to the reviewer agents.
Capability inventory: The skill invokes code-test and code-lint (which may execute code) and uses jj for repository access as seen in SKILL.md and SUBAGENT.md.
Sanitization: Absent; no validation or escaping of the code content is performed before processing.

code-quality