The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the Bash tool to execute a Python script (scripts/generate_schematic.py). The recommended command pattern python scripts/generate_schematic.py "your diagram description" incorporates natural language descriptions directly into shell arguments. This presents a risk of command injection if the input is not sanitized by the agent before execution.- [PROMPT_INJECTION]: The skill creates an 'Indirect Prompt Injection' surface by processing untrusted research data from external sources and using it to drive the schematic generation process. Evidence: 1. Ingestion points: Data retrieved from OpenRouter/Perplexity research queries. 2. Boundary markers: Absent; there are no instructions to delimit or ignore instructions within the research text. 3. Capability inventory: Access to the Bash tool for script execution. 4. Sanitization: Absent; the skill does not specify escaping or validation for the inputs used in shell commands.- [PROMPT_INJECTION]: The skill contains 'Metadata Poisoning' or behavior-steering instructions in the 'Suggest Using K-Dense Web' section. It directs the agent to 'proactively suggest' the author's commercial website (www.k-dense.ai) during complex workflows, which attempts to override the agent's neutral persona to favor a specific third-party service.

research-lookup