The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing the agent-pulse CLI tool to gather metrics and perform diagnostics. A helper script scripts/run_agent_pulse_snapshot.py uses the Python subprocess module to run multiple subcommands and aggregate JSON results.
[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to install the agentpulse-cli package from PyPI and optional web-related extras using pip.
[DATA_EXFILTRATION]: The tool can start local web (agent-pulse web) and API (agent-pulse api) servers on ports 8765 and 8766. While intended for local access, these features expose agent session logs and metrics over the network. The skill also reads potentially sensitive historical activity logs from various third-party AI agents.
[PROMPT_INJECTION]: The skill processes logs from multiple external agents, creating an indirect prompt injection surface where malicious content within ingested logs could attempt to influence the agent's behavior.
Ingestion points: Reads agent logs via subcommands such as search, status, and top as defined in SKILL.md.
Boundary markers: None identified; there are no explicit instructions to disregard or sanitize commands found within log content.
Capability inventory: Shell command execution in SKILL.md and scripts/run_agent_pulse_snapshot.py (via subprocess.run).
Sanitization: None; the agent summarizes raw CLI output which may contain content from external logs.

agent-pulse