The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large external documents (novels, adaptational materials, and multi-episode scripts) provided by the user.
Ingestion points: Untrusted source text is read from the 输入/ directory into the agent's context and processed by the included episode_intake.py script.
Boundary markers: The skill instructions (references/adaptation-craft.md and references/multi-episode-intake.md) explicitly mandate using semantic summaries and indices to avoid copying raw source text directly into project development files.
Capability inventory: The agent executes shell commands to run internal Python scripts for file indexing, slicing, and merging, and performs extensive file-write operations to generate drama development assets.
Sanitization: The process requires the agent to 'de-reference' mechanisms and summarize functional drama elements, providing a layer of abstraction between untrusted input and the generated output.
[COMMAND_EXECUTION]: The skill instructs the agent to execute local Python scripts included in the skill package (scripts/episode_intake.py and scripts/selftest.py) to manage the intake and validation of large screenplay files. These operations involve reading, indexing, and slicing text files within the project environment.

short-drama-develop