The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill's documentation in SKILL.md and references/here-now-publishing.md explicitly instructs the agent to run curl -fsSL https://here.now/install.sh | bash. This pattern is a critical security risk as it executes unverified code from a remote server with the user's local privileges.
[EXTERNAL_DOWNLOADS]: Setup scripts (scripts/setup-kokoro.sh and scripts/setup-piper.sh) download large binary models and executable components from external sources, including HuggingFace and third-party GitHub releases (github.com/thewh1teagle). These assets are utilized during the audio generation process and are not verified for integrity or safety.
[DATA_EXFILTRATION]: The core purpose of the skill is to extract content from local files or URLs and upload the resulting summaries to https://here.now. The scripts/publish.py script facilitates this by sending processed data to an external API, which could lead to sensitive information exposure.
[COMMAND_EXECUTION]: The skill frequently executes local shell scripts that perform environment configuration, such as creating virtual environments (venv) and installing packages (pip install). This modifies the host system's software environment.
[PROMPT_INJECTION]: The skill is designed to process untrusted data from external URLs and local files (e.g., in references/source-extraction.md). This creates an indirect prompt injection surface where malicious data could influence the agent's behavior.
Ingestion points: URLs, local files, and pasted text processed in SKILL.md (Step 2).
Boundary markers: The skill includes a 'Prompt-Injection Rule' in references/source-extraction.md to ignore source-embedded instructions, providing a basic defense.
Capability inventory: The skill can read local files, execute shell scripts, and perform network requests to upload data.
Sanitization: There is a revision pass mentioned in SKILL.md (Step 3) to ensure text is speakable, but no robust technical sanitization of the input data is present.

pipa-audio-brief