The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill configures monitoring surfaces that ingest external application data, including traces, logs, and frontend events. This data is untrusted and could potentially contain malicious instructions intended for the agent.
Ingestion points: OTel traces and span metrics (SKILL.md), Faro RUM sessions and events (references/faro.md), and LLM observability spans (references/ai-observability.md).
Boundary markers: None identified; instructions do not specify delimiters or instructions to ignore content within the ingested telemetry.
Capability inventory: The skill provides commands for running the alloy collector, using curl for verification, and managing packages via npm and pip (SKILL.md).
Sanitization: Setup steps do not include instructions for sanitizing or validating external telemetry data before ingestion into the observability platform.
[PRIVILEGE_ESCALATION]: Setup instructions include commands that require elevated privileges, such as the use of sudo to install the Grafana Alloy collector via the system package manager in references/apm-setup.md.
[EXTERNAL_DOWNLOADS]: Fetches official instrumentation binaries and SDKs from the OpenTelemetry and Grafana GitHub repositories and the unpkg CDN for frontend libraries.
[COMMAND_EXECUTION]: Employs standard CLI utilities to manage the observability infrastructure, including the alloy collector binary, curl for service health verification, and jq for parsing JSON output.

app-observability