The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs users to install a Python package named azure-ai-transcription in references/sdk/azure-ai-transcription-py.md. This package is not a recognized official Microsoft SDK and is not available on standard public registries like PyPI. The official package for these capabilities is typically azure-cognitiveservices-speech. Referencing non-existent or unverifiable packages introduces risks of dependency confusion and installation failures.
[PROMPT_INJECTION]: The skill's YAML frontmatter in SKILL.md lists the author as "Microsoft", which contradicts the vendor information provided for the skill ("attentiondotnet"). This misleading metadata may cause users to misjudge the skill's origin and safety.
[PROMPT_INJECTION]: The skill tools the agent to ingest and process external data, creating a surface for indirect prompt injection.
Ingestion points: Data enters the agent's context through search results via the azure__search tool and speech-to-text transcriptions via the azure__speech tool.
Boundary markers: The instructions do not require the agent to use delimiters (like triple backticks or XML tags) or to include "ignore embedded instructions" warnings when processing the retrieved external content.
Capability inventory: The skill provides tools for search, transcription, and various SDK operations that can process significant amounts of attacker-controlled content.
Sanitization: There are no instructions or automated checks in the skill to sanitize or validate external input before it is used to influence the agent's reasoning or actions.

azure-ai