The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local repository (such as README files, specification briefs, and existing source code) to generate architectural documentation and phased plans. This represents a potential surface for indirect prompt injection if the source files contain malicious instructions.
Ingestion points: The skill uses Grep, Glob, and an Explore subagent to scan directories and files listed in references/discovery.md (e.g., app/, pages/, package.json) and SKILL.md (briefs).
Boundary markers: No explicit delimiters or specific instructions are used to isolate ingested data from agent instructions during processing.
Capability inventory: The skill possesses Write, Edit, and Agent permissions to generate documentation artifacts, but it does not autonomously execute shell commands or interact with external networks.
Sanitization: The skill does not perform sanitization on the ingested codebase content before using it to generate documentation.
[SAFE]: No evidence of malicious code execution, data exfiltration, or obfuscation was detected. The skill's primary function is documentation and project planning for human review.

information-architecture