The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill indexes arbitrary files from local projects (code, SQL, specs, docs), which creates a vulnerability surface for indirect prompt injection where malicious content in those files could influence the agent's behavior during indexing or search tasks.
Ingestion points: Project files are read and processed by codebase_index, codebase_update, and codebase_context_index in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or boundary markers to isolate indexed data from agent instructions.
Capability inventory: The skill includes tools for data removal (codebase_remove, codebase_prune) and process termination (codebase_stop).
Sanitization: No sanitization or validation procedures for indexed content are mentioned in the documentation.
[COMMAND_EXECUTION]: The codebase_stop tool interacts with system processes by sending SIGTERM signals to orphan processes that hold file locks, as described in references/tool-reference.md.
[EXTERNAL_DOWNLOADS]: The skill documentation references software downloads from well-known services, including Docker Desktop (docker.com) and Ollama (ollama.com), for infrastructure setup and performance acceleration.

codebase-management