The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides shell scripts (scripts/mirror.sh, scripts/slice.sh) for site mirroring and media processing. These scripts execute system-level commands including curl, ffmpeg, and perl. Notably, scripts/mirror.sh lacks path validation when saving downloaded assets, which could potentially lead to path traversal if used against a malicious source.
[COMMAND_EXECUTION]: Step 8 of the workflow in SKILL.md instructs the agent to run a headless Google Chrome command locally to capture screenshots for quality assurance.
[EXTERNAL_DOWNLOADS]: The scripts/mirror.sh tool enables the downloading of web content and assets from user-defined external URLs. This functionality allows the agent to interact with and fetch data from arbitrary remote locations.
[INDIRECT_PROMPT_INJECTION]: The mirroring process (Step 0) introduces an attack surface for indirect prompt injection. By ingesting untrusted files (HTML, CSS, JS) from external websites, the agent may be exposed to malicious instructions that could influence its behavior during the subsequent engine generation and styling steps.
Ingestion points: scripts/mirror.sh downloads external files into the workspace.
Boundary markers: Absent; no delimiters or instructions to ignore embedded content are provided.
Capability inventory: The agent has access to Bash, Read, Write, and Edit tools.
Sanitization: Absent; content is extracted via regular expressions without semantic sanitization.

stopscroll-hero