The Agent Skills Directory

[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it processes untrusted data from source code and repository documentation (e.g., docs/ARCHITECTURE.md, CLAUDE.md). This is inherent to the skill's purpose as a code review tool.
Ingestion points: Reads source code, test files, and project documentation as specified in the context metadata and instructions.
Boundary markers: Not explicitly defined for external content.
Capability inventory: Writing review artifacts to .reviews/ and audit trails to .factory/audit-trail/.
Sanitization: No explicit sanitization of input code content before processing.
[COMMAND_EXECUTION]: The documentation includes a command to install a dependent skill (npx skills add jwilger/agent-skills --skill domain-modeling). This is a standard installation pattern for the author's own resources and does not involve untrusted third-party code.
[SAFE]: File system activity is restricted to local project directories for persisting review findings (.reviews/) and pipeline state (.factory/). These operations are consistent with the skill's stated purpose of providing durable review artifacts.

code-review