The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted text from PR comments and evaluation results which could contain malicious instructions designed to influence agent behavior.
Ingestion points: Review comments (in the 'Review Comment Triage' section) and RDE eval results (in the 'RDE Rule Validation' section).
Boundary markers: None identified.
Capability inventory: Shell command execution via bunx and nr (npm-run).
Sanitization: No validation or sanitization is specified for external input.
[COMMAND_EXECUTION]: The skill executes shell commands such as bunx @rayhanadev/truffler and nr changeset. The truffler command interpolates a helper name directly into the command string, which could lead to command injection if the helper name is derived from untrusted code without sanitization.
[EXTERNAL_DOWNLOADS]: The skill utilizes bunx to download and run the @rayhanadev/truffler package from the npm registry if it is not already available locally.

rule-validate