The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages files on the local filesystem, including writing project state to DISCOVERY.json and PROJECT.json. It also provides functions to delete existing project artifacts through a user-confirmed 'Restart Guard' process.
[DYNAMIC_EXECUTION]: The skill invokes a local JavaScript file (scripts/presentation-theme-invalidation.mjs) to determine which files should be invalidated or preserved during project updates. This script is used for internal path calculations based on changes to themes or fonts.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection via user-supplied interview answers.
Ingestion points: User responses to the discovery questions defined in QUESTIONS.md are captured and processed.
Boundary markers: The skill implements a confirmation step in SKILL.md (Step 2), where the agent presents a summary of all gathered data for user verification before any files are written.
Capability inventory: The skill has the capability to write configuration files (WRITE_DISCOVERY.md), delete project files (RESTART-GUARD.md), and execute a local helper script (RESTART-GUARD.md).
Sanitization: The skill structures the data into a JSON schema, which provides some implicit formatting, though it does not explicitly sanitize the text content provided by the user.

discover-presentation