The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user configuration data, such as OTLP endpoints, headers, and attributes, which are interpolated into proposed shell commands.
Ingestion points: User-provided values for telemetrygen CLI flags in SKILL.md.
Boundary markers: Instructions mandate shell-quoting of literal values and the use of environment placeholders for sensitive data.
Capability inventory: The skill constructs shell commands and includes a verification harness that uses Docker and Python.
Sanitization: Instructions require strict input validation, target authorization review for shared infrastructure, and the refusal of dangerous flag combinations like unbounded rates or durations.
[COMMAND_EXECUTION]: The references/collector-verification.md file contains a multi-step shell script for local collector testing. The script incorporates security checks, such as character validation for container names and absolute path enforcement, to mitigate potential command injection during the setup and cleanup process.
[DYNAMIC_EXECUTION]: The verification harness dynamically generates and executes a short Python script using a heredoc to process JSONL output files. This script uses standard Python libraries (json, pathlib, sys) and is based on a static template provided within the skill documentation.
[EXTERNAL_DOWNLOADS]: The skill provides instructions to fetch the telemetrygen binary and related Docker images from the official OpenTelemetry GitHub and GHCR repositories. These downloads originate from well-known services and trusted organization repositories.

otel-telemetrygen