CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

review-telemetry-privacy-perf

Review system telemetry collection, OS monitoring daemons, and client metric streams for privacy, data minimization (zero keystroke/PII logging, aggregate bandwidth only), CPU sampling overhead, and connection resilience.

Quality

80%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./skills/review-telemetry-privacy-perf/SKILL.md
SKILL.md
Quality
Evals
Security

Review Telemetry, Privacy & Performance

A specialized review lens for system telemetry collectors, OS metrics polling, and daemon streaming (src/telemetry/**). Ensures that ambient system monitoring respects user privacy, minimizes data footprint, and executes with negligible resource overhead.

1. Scope

Review only files responsible for gathering system statistics (CPU, memory, load, network I/O, battery power), event tracking (mouse, keyboard, GUI activity), and streaming telemetry payloads over WebSockets or IPC.

2. Review Rubric & Rules

A. Strict Data Minimization & Privacy (Must-Fix)

  • Zero Raw Input Capture: Never record, log, serialize, or transmit raw key codes, character strings, typed text, window titles, or clipboard contents.
  • Aggregate Network Throughput Only: Network telemetry must measure aggregate byte counters and transfer rates (bytesInPerSec, bytesOutPerSec). It must never inspect packet payloads, destination IP addresses, hostnames, or URLs.
  • Pulse/Rate Only: Activity metrics must be reduced strictly to numerical event counters, frequency deltas, or weighted activity pulse rates (e.g. activityRate: 0.0 - 1.0).
  • No sensitive user identifiers, host machine identities, or filesystem paths may be exposed in telemetry payloads.

B. Daemon CPU Overhead & Non-Blocking Sampling (Must-Fix)

  • Metric collection must not introduce noticeable CPU overhead. Sampling loops must be throttled (default interval >= 500ms).
  • CPU usage calculations must use delta diffing on cumulative CPU tick counters (user + sys + idle) rather than synchronous blocking sleep calls.
  • Sampling must not block the Node.js event loop or starve WebSocket message processing. Auxiliary I/O polling (such as network or power states) should be decoupled or run asynchronously.

C. Connection Lifecycle & Error Resilience (Consider)

  • Ensure WebSocket servers and client socket listeners gracefully handle unexpected disconnects, error events, and malformed client payloads without crashing the process.
  • Background intervals must be properly cleared on daemon termination (SIGINT/SIGTERM).

3. Reporting Guidance

  • Anchor each finding to the exact line number in the diff.
  • Flag any potential privacy leak as a high-severity / must-fix finding with clear remediation.
  • If all telemetry code adheres to data minimization and non-blocking performance standards, confirm approval succinctly.
Repository

colinhallett/tcr-ambient-monitor

Path

skills/review-telemetry-privacy-perf/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.