The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to invoke other local commands or skills, such as /create-verification-skill and /create-formal-verification-skill, based on user consent to assist with project verification.
[PERSISTENCE]: The skill writes a configuration file to ~/.cursor/rules/pstack-models.mdc with the alwaysApply: true property. This creates a persistent instruction set that the AI agent will automatically load and follow in future sessions within the Cursor IDE to maintain model preferences.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external files which could theoretically be manipulated to influence the agent's configuration logic.
Ingestion points: Reads current configuration from ~/.cursor/rules/pstack-models.mdc and mapping information from ../orca-harness/SKILL.md.
Boundary markers: No explicit markers or warnings are used when reading these files to prevent processing of embedded instructions.
Capability inventory: File system write access (overwriting configuration) and invocation of verification-related skills.
Sanitization: The skill performs validation in Step 4, ensuring that any model slugs written to the configuration are confirmed as available in the current session, mitigating the risk of injecting invalid or malicious model targets.
[DYNAMIC_EXECUTION]: The skill offers to generate new project-local verification skills (/create-verification-skill) to allow agents to drive the application and verify changes, which involves dynamic generation of executable skill files.

setup-pstack