The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for workspace initialization, git operations (submodule add, init, push), and running utility scripts (node, npm).\n- [EXTERNAL_DOWNLOADS]: The workflow/create-mode.md workflow executes npm install to download the pdf-parse library from the public npm registry during the workspace setup process.\n- [CREDENTIALS_UNSAFE]: As evidenced in evals/evals.json, the skill is designed to read authentication information from ~/.config/agent-wiki/credentials to facilitate remote repository access.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8) due to its core functionality.\n
Ingestion points: Reads unstructured data from external .pdf, .txt, and .md files provided by the user (as seen in workflow/create-mode.md).\n
Boundary markers: Absent. There are no instructions provided to the agent to treat file contents as untrusted or to use delimiters to prevent command injection via file text.\n
Capability inventory: The skill can execute shell commands, perform git operations, and modify the file system across multiple workflow steps.\n
Sanitization: Absent. The skill instructions do not specify any filtering or escaping of content extracted from source documents before integrating them into the AI-focused AGENTS.md or backlog files.

agent-wiki