The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes project-local configuration files including .pstack/config.md, .pi/settings.json, and opencode.json. While this creates a surface for indirect prompt injection if those files were to be compromised by an attacker, the risk is mitigated by the skill's logic, which requires concrete model IDs to be validated against the live host inventory.\n
Ingestion points: Reads from paths like $PSTACK_CONFIG, .pstack/config.md, .pi/settings.json, and opencode.json.\n
Boundary markers: None defined for the ingestion of existing configuration file content.\n
Capability inventory: The skill has the capability to write to the local filesystem and execute platform-specific CLI tools.\n
Sanitization: All model IDs written to configuration must match the output of the host's live inventory command (pi --list-models).\n- [EXTERNAL_DOWNLOADS]: The skill suggests and facilitates the installation of the pi-subagents package from the NPM registry to enable parallel worker and per-role model capabilities on the Pi platform.\n- [COMMAND_EXECUTION]: The skill uses host-level commands such as pi --list-models, subagent, and /subagents-doctor to perform environmental discovery and verify the presence of required delegation facilities.

setup-pstack