The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various opendeploy CLI commands to retrieve deployment status, service configurations, and logs. These commands are legitimate diagnostic operations for the OpenDeploy platform.
Evidence: opendeploy deployments logs, opendeploy services config get, opendeploy services config patch, and opendeploy context resolve in SKILL.md and reference files.
[COMMAND_EXECUTION]: Uses npm to verify the installation and version of the platform's official CLI package.
Evidence: npm list -g @opendeploydev/cli and npm view @opendeploydev/cli version in SKILL.md.
[DATA_EXFILTRATION]: The skill explicitly implements a redaction policy to prevent the accidental exposure of credentials or secrets while debugging.
Evidence: Hard rules in SKILL.md state "Do not print env values, API keys, bearer headers, bind signatures, decrypted secrets, or SSL private keys. Show key names only."
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from deployment and build logs. However, this risk is mitigated by strict instruction following and redaction rules.
Ingestion points: Deployment and service logs fetched via opendeploy ... logs (found in SKILL.md and references/logs.md).
Boundary markers: Not present.
Capability inventory: Configuration mutation via opendeploy services config patch (found in references/port.md and references/startup-order.md).
Sanitization: Instructions mandate the redaction of sensitive values and excerpts in the final report.

opendeploy-debug