The Agent Skills Directory

[METADATA_POISONING]: The skill description claims to provide "Anthropic's official brand colors and typography," which is misleading given the skill is provided by an external author ("huangzida"). Additionally, the LICENSE.txt file includes a copyright notice for "Anthropic, PBC" with a future date (2026), which is inconsistent and deceptive.
[INDIRECT_PROMPT_INJECTION]: The skill's stated purpose is to apply styling to "any sort of artifact," creating an attack surface where malicious instructions within the artifact could influence the agent during the styling process. * Ingestion points: Processes user-provided artifacts for brand styling (SKILL.md). * Boundary markers: No delimiters or "ignore" instructions are provided to separate the artifact's content from the agent's styling task. * Capability inventory: The skill specifies styling logic for text and shapes, referencing the use of the python-pptx library for these modifications (SKILL.md). * Sanitization: No input validation or sanitization of the artifact content is described.
[NO_CODE]: The skill consists entirely of informational guidelines and standard license text, with no executable scripts, binaries, or command-line tools included.

brand-guidelines