The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple bash commands to perform project analysis. This includes using find, awk, wc, sort, and uniq to calculate file counts, directory depth, and code concentration. These operations are core to its discovery phase.
[PROMPT_INJECTION]: The skill processes untrusted data from the analyzed repository, including existing AGENTS.md and CLAUDE.md files as well as source code symbols and exports. This ingestion of external data without explicit sanitization or strict boundary markers creates a surface for indirect prompt injection, where malicious instructions hidden in the codebase could influence the agent's behavior during documentation generation.
Ingestion points: Discovery phase reads existing AGENTS.md, CLAUDE.md, and project source files (via LSP, codegraph, and Read tool).
Boundary markers: The skill lacks explicit instructions to ignore embedded commands in the files it analyzes.
Capability inventory: The skill can execute bash commands, write/edit files, and spawn sub-agents with specific tasks.
Sanitization: There is no evidence of filtering or sanitizing external content before it is used to inform the agent's plan or generated output.

init-deep