The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill recommends installing software using a dangerous pattern that executes a remote script directly in the shell.
Evidence: curl -fsSL https://raw.githubusercontent.com/tsirysndr/otelview/main/install.sh | bash in SKILL.md.
[COMMAND_EXECUTION]: The skill provides multiple commands for the agent or user to execute which involve installing global packages and binary tools, posing a risk of arbitrary code execution.
Evidence: brew install tsirysndr/tap/otelview, bun install -g otelview, and manual MCP registration commands.
[EXTERNAL_DOWNLOADS]: The skill relies on external, unversioned resources for its core functionality and installation.
Evidence: References to a shell script on raw.githubusercontent.com and a custom Homebrew tap.
[INDIRECT_PROMPT_INJECTION]: The skill provides tools to ingest and display data from external, potentially untrusted sources (logs, traces, and metrics), which could contain malicious payloads.
Ingestion points: Tools such as search_logs, search_traces, and investigate_trace read data from a telemetry server or database file.
Boundary markers: None present; the skill does not instruct the agent to distinguish between telemetry data and command instructions.
Capability inventory: The skill possesses the ability to execute shell commands and perform network requests to an MCP endpoint.
Sanitization: There is no evidence of sanitization or filtering applied to the telemetry data before it is presented to the agent.

otelview