The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes a surface area for indirect prompt injection as it requires the agent to ingest and analyze output from external network requests and system deployment statuses.
Ingestion points: External data enters the agent's context through curl health check responses from staging/production endpoints and kubectl rollout/status logs as described in SKILL.md.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore potential commands embedded within the external data.
Capability inventory: The skill uses kubectl (for management and rollback) and curl (for network probing) as defined in the allowed-tools section and used in SKILL.md.
Sanitization: No specific mechanisms for filtering or validating the structure of external responses are described before the agent processes the information.
[COMMAND_EXECUTION]: The skill uses kubectl to perform deployment operations such as applying configurations, monitoring rollout status, and executing rollbacks. It also utilizes curl for connectivity and health verification. These commands are used legitimately for the skill's stated purpose.
[EXTERNAL_DOWNLOADS]: The skill references and fetches status information from status.cohere.com. This reference targets the official status page of a well-known AI service provider and is consistent with standard production monitoring practices.

cohere-prod-checklist