The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill automates the upload of generated reports to GitHub Gists using the gh command-line tool. While these gists are created as "secret," they are unlisted but remain publicly accessible to any party that obtains the raw URL.
[COMMAND_EXECUTION]: The pattern utilizes the GitHub CLI (gh) to create gists and query the GitHub API to construct preview links.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]:
The generated HTML reports load the Tailwind CSS framework from cdn.tailwindcss.com and the Mermaid diagramming library from cdn.jsdelivr.net. Both are well-known and established services.
The Mermaid library is initialized with securityLevel: "loose". This configuration disables the internal sandbox, which could permit the execution of arbitrary HTML or JavaScript if the data processed by the agent to create the diagrams contains malicious payloads.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests codebase architecture details, dependency graphs, and performance metrics from the local environment (SKILL.md).
Boundary markers: No specific delimiters or warnings are used to separate untrusted codebase data from the report generation instructions.
Capability inventory: The skill has the capability to write files to the docs/reports/ directory and transmit data to external servers via GitHub Gists.
Sanitization: The pattern does not include explicit steps for sanitizing or escaping codebase content before it is embedded into the HTML/Mermaid report templates.

publish-html-report