The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external source code and logs using file-reading tools and shell access, creating a potential surface for indirect prompt injection. The risk is minimized by the skill's restrictive instructions which limit the agent's responses to single questions during the investigative phases.
Ingestion points: Data enters the context through the Read, Grep, Glob, and Bash tools used to examine the user's code as instructed in SKILL.md.
Boundary markers: No specific delimiters or instruction-ignore warnings are defined for the analyzed content.
Capability inventory: The skill has access to shell execution (Bash), file system operations (Read, Grep, Glob), and can trigger secondary workflows (systematic-debugging).
Sanitization: No explicit filtering or sanitization of external file content is performed.
[SAFE]: The skill's instructions are focused on educational guidance and do not contain any patterns associated with data exfiltration, persistence, or privilege escalation. The workflow ensures user oversight by requiring confirmation before escalating to direct debugging.

socratic-debugging