The Agent Skills Directory

[COMMAND_EXECUTION]: The skill employs standard bash utilities including find, awk, sed, wc, and sort to perform structural analysis, calculate directory depth, and measure code concentration. These operations are scoped to the project environment and specifically exclude common non-source directories such as node_modules, venv, and .git.
[PROMPT_INJECTION]: The skill processes content from existing documentation files and source code symbols to generate new documentation, creating a surface for indirect prompt injection. \n
Ingestion points: Reads AGENTS.md and CLAUDE.md files; extracts symbols via LSP (lsp_document_symbols, lsp_workspace_symbols). \n
Boundary markers: Absent; sub-agent prompts do not include specific delimiters or instructions to ignore embedded content. \n
Capability inventory: Spawns sub-agents (Task) and has the ability to write files (Write). \n
Sanitization: Ingested content is merged into context without explicit escaping or validation.
[SAFE]: All detected behaviors are consistent with the skill's primary function of documentation generation. No credential harvesting, network exfiltration, persistence mechanisms, or unauthorized privilege escalations were identified.

index-knowledge