The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several external command-line interface (CLI) tools including claude, codex, grok, and spawn_agent. These are used to probe model availability and verify credentials as part of the setup process.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external configuration files which could theoretically be manipulated to influence the setup logic.
Ingestion points: Reads and processes pstack-models.md, CLAUDE.md, and AGENTS.md (specifically bounded blocks).
Boundary markers: Uses HTML comment markers ( and ) to delimit configuration blocks in AGENTS.md.
Capability inventory: Includes the ability to write to system configuration files and execute shell commands through multiple external launchers.
Sanitization: Implements normalization for legacy model aliases and validation rules against a defined model matrix to ensure only supported descriptors are written.
[PERSISTENCE]: The skill modifies the agent's primary configuration files (CLAUDE.md or AGENTS.md) to integrate the pstack-models.md file. While this is the intended purpose of the configuration utility, these files persist across sessions and dictate future agent behavior.

setup-pstack