The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the agg binary from the official asciinema GitHub repository. It demonstrates good security posture by requiring a specific version and performing a SHA256 checksum verification before installation.
[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to manage build processes (cargo), terminal sessions (tmux), and media transcoding (ffmpeg, asciinema). It also requires sudo privileges to install dependencies and the agg tool.
[CREDENTIALS_UNSAFE]: The instructions involve the use of WARP_API_KEY for cloud-based verification. The skill explicitly warns the agent not to echo, print, or inline the secret value, relying instead on inherited environment variables to mitigate the risk of credential exposure.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads the contents of a terminal pane using tmux capture-pane. If the TUI renders untrusted data, the agent could interpret that data as instructions.
Ingestion points: The agent reads the rendered screen text via tmux capture-pane -t tuicheck -p in SKILL.md.
Boundary markers: Absent; the skill does not specify the use of delimiters when processing the captured terminal text.
Capability inventory: The skill has access to cargo build, tmux control, and package management via apt-get as described in SKILL.md.
Sanitization: Absent; there is no logic to filter or escape instructions that might be embedded in the terminal buffer.

tui-verify-change