The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the user's repository which could contain malicious instructions designed to influence the agent.
Ingestion points: Reads local source code files, git log output, CONTEXT.md, and architectural decision records in docs/adr/ as noted in SKILL.md.
Boundary markers: No explicit delimiters or instructions to ignore embedded prompts within the ingested codebase data are provided to the agent.
Capability inventory: The skill can execute shell commands, write files to the temporary directory, and spawn specialized sub-agents for exploration and modeling.
Sanitization: No sanitization or escaping of ingested file content is performed before it is used in report generation or sub-agent prompting.
[COMMAND_EXECUTION]: The skill executes local shell commands to gather project data and display report files.
Evidence: Uses git log --oneline to identify hotspots in the code (SKILL.md).
Evidence: Uses system utilities xdg-open, open, or start to open generated HTML reports in a web browser (SKILL.md).
[EXTERNAL_DOWNLOADS]: The generated HTML documentation fetches resources from well-known content delivery networks.
Evidence: Loads the Tailwind CSS library from cdn.tailwindcss.com (HTML-REPORT.md).
Evidence: Loads the Mermaid.js library for diagramming from cdn.jsdelivr.net (HTML-REPORT.md).
Context: These are well-known services used for stylistic rendering and do not contribute to verdict escalation under standard service recognition rules.

vibe-refactor