The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command open dashboard.html to launch the generated monitoring interface in the user's default browser.
[DYNAMIC_EXECUTION]: The skill generates a self-contained HTML file containing inline JavaScript and CSS logic. This script is designed to fetch and process results.json data to dynamically update a progress dashboard.
[INDIRECT_PROMPT_INJECTION]: The skill has a high-exposure surface for indirect prompt injection. It ingests the content of other skills and user-provided test inputs, then processes the output of those skills to drive a mutation loop. It specifically instructs the agent to "Analyze failures" and rewrite the target SKILL.md file based on those failures.
Ingestion points: Target SKILL.md files, user-supplied test inputs, and outputs from the executed target skill.
Boundary markers: No specific delimiters or safety instructions are defined to separate the mutated prompt from the optimization instructions.
Capability inventory: File system write access (SKILL.md, results.json, results.tsv), shell execution (open).
Sanitization: No explicit sanitization of the target skill content or its output is performed before it is used to modify the source code of the skill.
[EXTERNAL_DOWNLOADS]: The generated dashboard is instructed to fetch the Chart.js library from a public CDN (jsdelivr.net). While this targets a well-known service, it introduces a remote dependency into a locally generated executable file.

autoresearch