The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads arbitrary files from the codebase (untrusted data) and passes their content or structure to sub-agents for analysis and documentation generation. A maliciously crafted comment or code pattern could potentially influence the sub-agents' behavior or the final documentation output.
Ingestion points: Codebase files are explored and read via shell commands, LSP symbol analysis, and sub-agent analysis tasks.
Boundary markers: No delimiters or "ignore embedded instructions" warnings are used when processing the content of project files.
Capability inventory: The skill has access to shell execution (bash), LSP interaction, sub-agent spawning (Task), and file writing.
Sanitization: No sanitization or validation of the ingested code content is performed before generating the AGENTS.md files.
[COMMAND_EXECUTION]: The skill uses shell commands to analyze the project's structure, file counts, and code concentration.
Evidence: Execution of commands like find, awk, wc, sed, and sort to calculate project scale and directory depth.

index-knowledge