The Agent Skills Directory

[METADATA_POISONING]: The skill metadata claims the author is 'Google' and states it is part of the 'Google ADK skills suite'. This directly contradicts the author identified in the system context ('oscarrondon-dev'). This deceptive branding could lead users to trust the skill's instructions and tool requirements based on the reputation of a trusted organization.
[REMOTE_CODE_EXECUTION]: The skill instructs the user to install a package named google-agents-cli via uv tool install. Due to the discrepancy in authorship, this represents an unverifiable dependency from an unconfirmed source, posing a risk of malicious package execution under a misleading name.
[EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch documentation and integration details from adk.dev. This domain is not recognized as a trusted organization or well-known service, and its relationship to the official tool suite is unverified.
[INDIRECT_PROMPT_INJECTION]: The skill relies on external content from an untrusted source, which could be used to influence agent behavior.
Ingestion points: Multiple documentation URLs (e.g., https://adk.dev/observability/index.md) fetched by the agent.
Boundary markers: None. The skill does not instruct the agent to sanitize or ignore instructions embedded in the external documentation.
Capability inventory: Access to powerful CLI tools (agents-cli, bq, gsutil) for cloud infrastructure management and data querying.
Sanitization: None; fetched content is processed directly to guide the agent's interactions.

google-agents-cli-observability