The Agent Skills Directory

[COMMAND_EXECUTION] (HIGH): The SKILL.md file contains a 'dependency self-healing' instruction that commands the agent to 'immediately auto-execute' pip install -r [SKILL_PATH]/requirements.txt without asking the user if an ImportError is encountered. This bypasses user oversight for shell command execution.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill identifies and installs external Python dependencies at runtime. While the packages in requirements.txt appear standard, the automated installation process facilitates potential supply chain risks or unauthorized software installation.
[PROMPT_INJECTION] (LOW): Indirect Prompt Injection Surface. The skill is designed to ingest and 'deeply analyze' content from arbitrary URLs or local files, which is then used to influence the agent's subsequent 'Master Persona' output.
Ingestion points: scripts/content_ingester.py processes external inputs; data is stored in config/raw_content.txt.
Boundary markers: Absent. The skill does not define clear delimiters or instructions to the LLM to ignore potentially malicious commands embedded in the ingested documents.
Capability inventory: The skill can execute Python scripts, perform web searches via WebSearch, and write HTML/Markdown files to the local filesystem.
Sanitization: While html2text is used to clean HTML noise, no security-focused sanitization is performed on the ingested text before it is processed by the model.

knowledge-absorber