The Agent Skills Directory

[SAFE]: The skill is strictly read-only and intended for codebase analysis. It includes explicit instructions to avoid changing code, starting servers, or running destructive commands.
[COMMAND_EXECUTION]: The skill utilizes the standard command-line tool ripgrep (rg) for codebase exploration, which is consistent with its stated purpose and limited to local read access.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it processes untrusted data from the codebase being analyzed. 1. Ingestion points: Files read through the rg tool and direct file access as defined in SKILL.md and explorer-prompt.md. 2. Boundary markers: Absent. The prompts do not use delimiters or instructions to ignore embedded commands in the analyzed code. 3. Capability inventory: Restricted to read-only search and file reading; no network, file-write, or execution capabilities are present. 4. Sanitization: Absent. Data from the codebase is processed directly to generate explanations. The risk is minimized by the agent's restricted toolset.

how