The Agent Skills Directory

[SAFE]: The skill implements a structured 3-layer framework (Acquisition, Processing, Management) to guide the agent in analyzing codebases efficiently.
[COMMAND_EXECUTION]: Includes routine commands for environment inspection, such as ls -la and tree, which are used solely for understanding directory structures as part of the code reading process.
[DATA_EXPOSURE]: While documentation templates mention sensitive terms like JWT_SECRET or TokenValidator, these are strictly educational placeholders intended to show users how to record their own findings, not instructions to exfiltrate real secrets.
[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to process and interpret external code files. While this creates a potential surface for indirect prompt injection (if the code being read contains malicious instructions), the skill itself does not contain any logic that facilitates this or bypasses safety guardrails.

code-reading