The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes instructions for the agent to use a local Python script to navigate and retrieve specific sections from its large reference library.
Evidence: python scripts/read_sections.py references/style_patterns.md --list --find "Output Modes" mentioned in SKILL.md.
Analysis: The provided script (scripts/read_sections.py) is a utility that uses only Python standard libraries (argparse, json, pathlib, re, sys) to parse Markdown headings. It contains specific security checks, such as verifying that requested files are relative to the skill root and have a .md suffix, preventing path traversal or the reading of sensitive system files. This is a functional developer tool for context management.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted user content, such as novel excerpts or scene ideas, for the purpose of cinematic adaptation.
Ingestion points: Plot summaries, novel excerpts, or scene ideas provided by the user (as described in SKILL.md).
Boundary markers: The skill uses a structured 'Cinematic Translation' framework that forces the agent to extract only filmable visual/audible elements, which naturally mitigates the risk of the agent obeying hidden instructions that are non-visual in nature.
Capability inventory: Primary capability is text generation; secondary is local file reading via the provided Python script.
Sanitization: The 'Director Delivery Gate' and 'Physical-feasibility audit' requirements ensure the agent evaluates the input through a rigid cinematic lens before generating output.

cinematic-video-prompt-engineer