The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from external data sources processed at runtime.
Ingestion points: The skill fetches plugin definitions from the rhdh-plugin-export-overlays repository and monitors PR comments for responses from the @rhdh-test-bot (SKILL.md).
Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the fetched YAML data or monitored bot comments.
Capability inventory: The skill employs powerful CLI tools including oc, kubectl, vault, and gh, which could be targeted if external input sources are compromised.
Sanitization: No sanitization or validation of the retrieved content is performed before the agent processes or displays the data.
[COMMAND_EXECUTION]: Executes commands that handle high-privilege cluster administrator credentials.
The skill provides instructions to retrieve CLUSTER_ADMIN_USERNAME and CLUSTER_ADMIN_PASSWORD from HashiCorp Vault to perform authenticated operations on OpenShift clusters.
While it includes critical warnings to prevent secrets from appearing in tool output, the automated orchestration of these credentials by an AI agent remains a sensitive operation.
[EXTERNAL_DOWNLOADS]: Downloads configuration data from vendor-owned external sources.
The skill uses curl to fetch plugin catalogs and package metadata from the redhat-developer/rhdh-plugin-export-overlays repository on GitHub. These resources are from the official vendor infrastructure.

deploy