The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill is designed to fetch runbook steps from an external API (api.openclaw.com) and execute them locally as automated scripts or API calls.
Evidence: 'Execute automated steps from runbooks, such as running scripts or API calls' and the POST endpoint https://api.openclaw.com/sre-runbooks/v1/runbooks/{id}/execute.
[COMMAND_EXECUTION]: The skill uses a CLI tool (sre-cli) and a programmatic library to run system-level commands and maintenance procedures.
Evidence: CLI commands like sre-cli runbook get and Python calls like sre_runbooks.execute(id=123).
[EXTERNAL_DOWNLOADS]: The documentation instructs users to install external packages and SDKs from public registries that are not part of the trusted vendors list.
Evidence: Instructions to run npm install sre-runbooks-sdk and use the sre_runbooks Python module.
[DATA_EXFILTRATION]: The skill transmits system environment parameters and configuration overrides to a remote endpoint.
Evidence: POST requests to api.openclaw.com containing body data such as {"step": 1, "params": {"environment": "prod"}}.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted data from monitoring tools and external API responses to trigger actions.
Ingestion points: External runbook content from api.openclaw.com and alerts from integrated monitoring tools.
Boundary markers: No boundary markers or 'ignore' instructions are defined to separate untrusted data from the execution logic.
Capability inventory: Capability to execute scripts and system commands via sre-cli and the SDK execute function.
Sanitization: There is no mention of sanitization or validation for the content retrieved from external sources before execution.

sre-runbooks