The Agent Skills Directory

[PROMPT_INJECTION]: The skill's reference documentation (including files such as references/architecture-patterns.md and references/security-best-practices.md) is populated with over 1,200 sections of repetitive, auto-generated technical filler. This constitutes metadata poisoning and obfuscation by volume, as it misrepresents the complexity and utility of the skill. Such excessive content can exhaust the context window or attention of AI agents, potentially hiding actual instructions or causing processing failures.
[COMMAND_EXECUTION]: The SKILL.md identifies the agent's ability to apply infrastructure manifests to a Kubernetes cluster using the kubectl tool. While this is consistent with the skill's stated purpose for DevOps and control plane management, it grants the agent administrative capabilities over cloud infrastructure which could be abused if the agent is misled by the deceptive documentation.
[EXTERNAL_DOWNLOADS]: The documentation and examples reference external packages and OCI images hosted on xpkg.upbound.io, the official Crossplane package registry. These references are consistent with legitimate ecosystem usage and point to an established industry service.

devops/crossplane