The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data (HTML prototypes, PRD documents, and user-provided specifications) to generate 'Validation Contracts' and technical specifications.
Ingestion points: The skill reads repository configuration files (AGENTS.md, CLAUDE.md), git log, and external documents including HTML files, PRDs (Notion/PDF/Markdown), and Figma-related descriptions in Phase A1.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the ingested documents are provided to the agent.
Capability inventory: The skill has the capability to write files to the local repository (docs/specs/), execute local shell commands (git rev-parse, git log, open), and dispatch external analysis tools (playground:playground).
Sanitization: The skill lacks validation or sanitization of content extracted from untrusted inputs before interpolating it into specifications.
Downstream impact: The generated validation-contract.md is specifically designed to be consumed by other agentic tools (like test-designer and deep-review), creating a multi-step execution chain where malicious instructions in a PRD could influence the generation of tests or audit results.

spec-design