The Agent Skills Directory

[COMMAND_EXECUTION]: The file references/workflow-smoke.md contains various shell script snippets (using git, mktemp, python3, node) designed for local environment setup and testing of the woostack skills. These commands are intended for use by developers within a controlled, local repository context.
[DYNAMIC_EXECUTION]: The testing recipes in references/workflow-smoke.md utilize node -e to execute inline JavaScript for validating skill assets. This code interacts with local project files (shared/sync.mjs, apps/docs/scripts/skill-assets.mjs) to ensure configuration consistency and does not involve untrusted remote input.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a command router, ingesting user requests to delegate tasks to specific specialized skills. While this constitutes an ingestion surface, the skill implements no autonomous logic that would execute untrusted instructions, relying instead on the established boundaries of the sub-skills it references.
Ingestion points: User-provided intents and requests parsed in SKILL.md to determine appropriate command routing.
Boundary markers: The skill documentation emphasizes adherence to project-level rules in AGENTS.md and requires explicit user authorization for most operations.
Capability inventory: The skill primarily performs command index lookups and provides documentation; shell and file system operations are limited to the development-focused testing recipes.
Sanitization: Not explicitly implemented in the index itself, as it delegates actual task implementation to other skills with their own security constraints.

using-woostack