The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to run profilecli, pprof, and git. These commands are used as intended for retrieving performance profiles, analyzing stack traces, and verifying repository state.
[EXTERNAL_DOWNLOADS]: The instructions direct the user to download the profilecli binary from the official Grafana GitHub repository (https://github.com/grafana/pyroscope/releases/latest/download/). This is a trusted source for the vendor's own tools.
[INDIRECT_PROMPT_INJECTION]: The skill processes JSON and text output from a remote Pyroscope server. As this data is ingested into the agent's context without specific boundary markers or sanitization, there is a theoretical surface for indirect prompt injection if a malicious server provides crafted profile metadata or function names.
Ingestion points: Data retrieved from profilecli query series and profilecli query profile (SKILL.md).
Boundary markers: None identified; data is parsed and displayed directly.
Capability inventory: The agent can read files, search with grep, and execute profiling/git commands.
Sanitization: No explicit sanitization or validation of the remote server's output is performed before presentation or analysis.

profilecli-insights