The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill facilitates the download and execution of the Tesseract CLI and installation scripts (install.sh, install.ps1) from the vendor's GitHub repository (mirage-hq/tesseract). These are vendor-owned resources used to provide the skill's primary functionality.
[COMMAND_EXECUTION]: Included Python helper scripts (tesseract_media.py, tesseract_sound.py) execute shell commands using subprocess.run to interface with local media tools like FFmpeg and ffprobe for audio synthesis, loudness measurement, and media inspection.
[INDIRECT_PROMPT_INJECTION]: The skill implements a "Critic Review" workflow where sub-agents evaluate deliverables based on untrusted user input provided in a brief.md file.
Ingestion points: Untrusted user requests are ingested into brief.md and read by critic agents (references/critic-review.md).
Boundary markers: Critic agents are instructed to read the brief first, but no explicit delimiters or instructions to ignore embedded malicious commands are present.
Capability inventory: The main agent possesses capabilities for file system modification (tsrct commit) and command execution (ffmpeg), while the critics are instructed to avoid project-mutating commands.
Sanitization: No sanitization or validation of the user-provided brief is documented before ingestion by the sub-agents.
[DYNAMIC_EXECUTION]: The skill enables the authoring and execution of custom fragment shaders using WGSL (WebGPU Shading Language) for visual effects (references/custom-shader.md). These shaders are compiled and executed by the Tesseract rendering engine at runtime.

tesseract-3d