ollygarden-otel-collector-config-validation
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading and executing container images from the official OpenTelemetry repository on Docker Hub. This is documented as safe behavior because it targets a well-known service and incorporates mandatory security controls.
- The skill uses the official 'docker.io/otel/opentelemetry-collector-contrib' image.
- It requires the use of pinned SHA256 digests to prevent tampering.
- It enforces a requirement for explicit user consent before any image pull or container run occurs.- [REMOTE_CODE_EXECUTION]: The workflow involves running a remote container image to perform behavioral validation.
- The execution is gated by mandatory user confirmation.
- The skill uses a specific, immutable image reference to ensure execution environment integrity.- [COMMAND_EXECUTION]: The skill includes a pre-defined shell script template for running the validation harness locally using Docker or Podman.
- The script uses 'mktemp -d' to create isolated scratch directories for each test run.
- It binds internal services to the host loopback interface (127.0.0.1) to prevent external network exposure.
- It manages SELinux relabeling (:Z) only on specific temporary files and output directories.
- It includes an EXIT trap to ensure the test container is removed after the validation is complete.- [INDIRECT_PROMPT_INJECTION]: The skill design involves processing user-provided configuration fragments which may contain untrusted data.
- Ingestion points: Production OpenTelemetry Collector configuration fragments provided by the user in SKILL.md and referenced in evals/evals.json.
- Boundary markers: The instructions guide the agent to wrap components in a specific harness template (references/harnesses.md) and isolate them from production ingress/egress.
- Capability inventory: The skill uses container runtimes to execute the provided configuration (references/run-harness.md).
- Sanitization: The skill relies on the isolation of the container runtime and the user's manual review of the validation plan rather than programmatic sanitization of the configuration content.
Audit Metadata