The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes a CLI utility named aiox to manage the development cycle. Commands include aiox sdc plan, aiox sdc next, aiox sdc verify, and aiox sdc status. These are used for state management and task orchestration.
[COMMAND_EXECUTION]: A mandatory safety gate aiox sdc preflight is executed before any subagent or model invocation. This command validates the intent, budget, and context files, halting execution on a non-zero exit code to prevent unauthorized model usage.
[DATA_EXPOSURE]: The orchestrator materializes prompt and context data into isolated local files before subagent dispatch. This is part of the governance process to ensure the data passed to the model matches the data validated by the preflight check.
[PROMPT_INJECTION]: The skill processes external content via the {story-path} argument. While this presents a surface for indirect prompt injection, the skill mitigates this risk by requiring a mechanical validation gate (preflight) and an explicit budget ceiling (AIOX_MODEL_BUDGET_CEILING_USD) before processing data through a model.
Ingestion points: Reads from the path provided in {story-path} (SKILL.md).
Boundary markers: None explicitly defined in the orchestrator instructions, relying instead on the external preflight validator.
Capability inventory: Executes shell commands via the aiox CLI and spawns subagents (aiox-po, aiox-dev, aiox-qa).
Sanitization: Performed externally by the aiox sdc preflight command which inspects the intent and context files.

full-sdc