The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from service logs and alarm history which could potentially contain malicious instructions.
Ingestion points: Data entering through opendeploy services logs and opendeploy monitoring alarms history commands.
Boundary markers: Uses AskUserQuestion to confirm actions with the user before executing mutations or engaging support.
Capability inventory: Access to Bash for CLI operations and AskUserQuestion for user interaction.
Sanitization: Instructions explicitly mandate that the agent 'never print secrets' and 'redact' evidence packets before any transmission.
[COMMAND_EXECUTION]: Executes the opendeploy CLI and npm for version management and alarm lifecycle operations. These tools are associated with the vendor resource patterns for the skill author 'opendeploy-dev'.
[DATA_EXPOSURE_AND_EXFILTRATION]: Transmits redacted evidence packets to OpenDeploy support channels. This is a primary feature for incident resolution and includes strict human-in-the-loop requirements for consent and evidence review.

opendeploy-alarms