The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external CSV files containing practice questions, answers, and explanations to generate video content and social media metadata.
Ingestion points: CSV data sources described in SKILL.md and references/bulk_csv_reels.md that are used for batch reel generation.
Boundary markers: The instructions lack explicit guidance for the agent to use boundary markers or "ignore" directives when interpolating untrusted CSV row content into generation prompts.
Capability inventory: The skill possesses significant capabilities including the execution of shell commands (manim, ffmpeg) and the creation of scripts that interact with the local file system.
Sanitization: While the documentation mentions generating "safe question slugs" for filenames, there are no specific requirements to sanitize or escape CSV content before it is processed by the DSPy-based generation agent.
[DYNAMIC_EXECUTION]: The workflow involves the automated generation of Manim Python scripts which are subsequently executed to produce video artifacts.
Evidence: SKILL.md (Step 5/6) and references/dspy_reelsgen_agent.md describe a pipeline where the agent writes scene logic to .py files and renders them using the manim CLI tool.
[COMMAND_EXECUTION]: The skill requires the execution of multiple command-line utilities for video production and verification.
Evidence: Routine use of manim, ffmpeg, ffprobe, and python commands for rendering, stitching media, and checking output specifications.
[EXTERNAL_DOWNLOADS]: The skill documentation refers to external libraries and local tools required for operation.
Evidence: Instructions to install manim via pip and the requirement for a local clone of the Qwen3-TTS repository at ~/clawd/Qwen-tts-voice-clone.

manim-reels