The Agent Skills Directory

Command Execution: The skill instructs the agent to run various command-line tools including cmake, ninja, ctest, and project-specific binaries like ./build/tools/rego. It also refers to using /tmp/opa for reference implementation testing. While these are standard in a developer environment, they involve executing code locally and should be used with trusted binaries.
Indirect Prompt Injection Surface: The development workflow involves reading and analyzing external Rego source files and OPA-generated JSON IR files. These external data sources represent a surface where maliciously crafted policy files could potentially influence agent behavior during analysis.
Ingestion points: Policy files (.rego) and OPA intermediate representation files (plan.json) read during the analysis phase.
Boundary markers: The skill does not explicitly define sanitization or boundary markers to isolate external file content from agent instructions.
Capability inventory: The skill allows for local file system writes and the execution of build and test utilities via the shell.
Sanitization: The workflow relies on standard compiler well-formedness checks but does not detail specific sanitization for external policy text before it enters the agent context.

trieste-dev