The Agent Skills Directory

[COMMAND_EXECUTION]: The skill manages the 'otelcol-contrib' Windows service using PowerShell (Stop-Service and Start-Service) as part of its log-pruning routine. While these operations are documented as part of the telemetry management workflow, they represent high-privilege interactions with host services.
[COMMAND_EXECUTION]: Pruning logic in otel/prune-otel-store.sh and otel/prune-collector-lifecycle.sh implements 'test seams' that allow overriding the service management commands via environment variables (e.g., CC_OTEL_STOP_CMD, CC_OTEL_START_CMD). This design presents a command injection risk if an attacker is able to influence the environment variables in the agent's shell session.
[COMMAND_EXECUTION]: The SKILL.md file uses dynamic context injection (the !command syntax) to execute shell commands like git, wc, and custom probes upon skill initialization to populate status summaries.
[EXTERNAL_DOWNLOADS]: The skill invokes npx -y ccusage to generate token usage reports. The ccusage package is an external dependency provided by the vendor 'melodic-software'.
[DATA_EXFILTRATION]: The skill processes sensitive local telemetry data that may contain prompt fragments or token counts. It addresses this risk with a comprehensive privacy filter and redaction layer defined in context/privacy.md which masks credentials, long Base64 strings, and other sensitive patterns before generating output.
[SAFE]: A static analysis detection of a reverse shell in otel/net-probe.sh is a false positive. The code uses bash's /dev/tcp feature as a legitimate, non-persistent probe to check if a local port is listening, immediately closing the file descriptor after the check.

observability