The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill dynamically generates a dashboard.html file containing inline JavaScript and CSS based on data processed during its execution loop.
[COMMAND_EXECUTION]: The skill uses the open command to execute and launch the generated HTML dashboard in the default web browser.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-provided test inputs, evaluation criteria, and outputs from the target skill, creating a vulnerability surface where malicious content could influence prompt mutations or generated dashboard content.
Ingestion points: User-provided inputs (test cases and eval criteria) and generated output of the target skill.
Boundary markers: None identified in the prompt mutation or dashboard generation steps.
Capability inventory: File system read/write for SKILL.md and logging files, and system command execution via open.
Sanitization: The skill lacks explicit sanitization for data incorporated into prompts or the dashboard.
[PROMPT_INJECTION]: The instructions mandate high autonomy, specifically directing the agent to run the experimentation loop without stopping for user confirmation and to autonomously rewrite instructions in other skill files.
[EXTERNAL_DOWNLOADS]: The generated dashboard is configured to fetch the Chart.js library from a public Content Delivery Network (CDN) when viewed in a browser.

autoresearch