The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted content from the codebase and documentation to generate design files, creating a surface for indirect prompt injection.\n- Ingestion points: Context is gathered from docs/<feature-name>/spec.md, docs/index.json, docs/architecture.md, and various source code files during context gathering as described in SKILL.md.\n- Boundary markers: The skill does not employ boundary markers or specific delimiters to isolate ingested file content from its primary instructions.\n- Capability inventory: Performs file read and write operations within the project workspace, specifically under the docs/ directory; no network or shell execution tools are enabled in the instructions or frontmatter.\n- Sanitization: Input content from specification files and source code is not sanitized or validated before being processed by the agent.

simple-design