The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest untrusted data from design documents, specifications, and existing source code to generate UI prototypes. While this presents a surface for indirect prompt injection, the risks are limited by the static nature of the output and the absence of high-privilege tools in the generation path.\n
Ingestion points: Processes plans, specifications, and code snippets as described in the brief extraction section of SKILL.md.\n
Boundary markers: No specific delimiters or boundary instructions are used when interpolating external content into the agent's context.\n
Capability inventory: The skill's operations are restricted to file system writes (index.html, prototype.css, prototype.js) and repository management (git mv).\n
Sanitization: There is no evidence of sanitization or filtering for the ingested content before it is used to generate executable static assets.\n- [EXTERNAL_DOWNLOADS]: A maintenance script is provided to check for updates to vendored documentation.\n
Evidence: The scripts/check-upstream script uses urllib.request.urlopen to download reference files from GitHub repositories for byte-for-byte comparison against local copies.\n
Note: This script is explicitly documented as a maintainer tool and is not part of the skill's runtime execution. The downloaded content is used for auditing purposes and is not executed.

qstack-ui-prototype