The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local script using the bun runtime. The script path is resolved dynamically via cc-plugin-root, which is a platform-specific command for locating plugin resources. This is a standard and expected mechanism for modular agent skills.
[DATA_EXFILTRATION]: The execution command explicitly unsets proxy environment variables (env -u HTTPS_PROXY -u HTTP_PROXY). This is likely done to ensure the script communicates directly with local or specific internal services without interference from system-wide proxy settings. No unauthorized external network requests or credential harvesting patterns were found.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a file (report.txt) to render its content into an image. While processing external text always presents a surface for indirect prompt injection, the skill's primary purpose is static rendering. The lack of automated decision-making based on the report's content significantly mitigates this risk.
Ingestion points: Reads from report.txt via command line arguments or standard input.
Boundary markers: None explicitly defined in the execution example, though the renderer uses specific prefix characters (#, !, >, .) for styling.
Capability inventory: File reading, image rendering via pushover_core.ts, and shell execution of the bun runtime.
Sanitization: The text is processed by a monospace renderer (@resvg/resvg-js) which treats the content as display text rather than executable instructions.

render-incident-report-image