openclaw - 💡(How to fix) Fix [Feature]: Skill usage telemetry — track which skills are actually being used [2 pull requests]

Problem

Currently there is no built-in way to track which skills have been actually used (CLIs executed) versus merely being discovered/scanned. The session-logs skill can search transcripts, but distinguishing between:

• System startup scanning SKILL.md files
• Heartbeat/auto exploration loading skills
• Actual user-triggered CLI invocation (the real usage)

requires manual grep and is error-prone.

Feature Request

Add a skill usage telemetry mechanism at the system level that:

1. Track skill invocations

Record when a skill's CLI tool is actually executed (not just SKILL.md read), with:

• Skill name
• Agent id
• Timestamp
• Command executed
• Success/failure

2. Expose usage data

• CLI: openclaw skills usage [--since 7d] [--agent clawbot9] → show usage count per skill
• UI/TUI: skills panel showing usage frequency, last used time
• /status or /skills command showing per-agent skill stats

3. Output format

Skill       Agent     Last Used    Count    Status
gog         main      2h ago       12       ✅
weather     clawbot9  1d ago       5        ✅
sonoscli    main      never        0        ⚠️ (never used)
imsg        clawbot9  5d ago       3        ✅

Use Case

Users have 41+ skills installed. Most are never used because:

• They don't know what's available or what's already working
• Skills with broken auth/CLI never get attention
• No visibility into which skills bring actual value

Analytics would help users:

1. Identify unused skills → clean up
2. See which skills are frequently used → prioritize further optimization
3. Detect skills that always fail → flag for fixing

Suggested Implementation

• Add a lightweight JSON log file per agent (e.g., ~/.openclaw/agents/{id}/skill-usage.json)
• Log on each exec call that matches a known skill binary
• Provide aggregation via session-logs skill or a new CLI subcommand

Environment

• OpenClaw version: 2026.5.18 (50a2481)
• OS: macOS 25.4.0 (arm64)
• Node: v24.12.0