Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →vivekchand/clawmetry-selfcheck
Read your own agent telemetry from ClawMetry (waste, progress, cost) and act on it before finishing a task.
Passed 12 of 18 test cases with the skill, and 7 of 18 without it.
and 7 more in the full report
8 fixed7 passed either way5 still failing2 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 8 test cases the skill fixed. Same model, same prompt, with and without the skill.
To investigate what actually happened in session SID-98234, you need the full session transcript. What subcommand and flag retrieves the transcript for a specific session ID in the agent telemetry CLI?
| Checked | Without skill | With skill |
|---|---|---|
| Specifies the --transcript flag with the session ID. | Missed | Passed |
| Retrieves the event log transcript for the designated session. | Passed | Passed |
The instructions your model reads when the skill loads.
ClawMetry observes AI agent sessions on this machine (read-only). Use its CLI to see your own telemetry and improve your behaviour mid-task.
All commands are read-only, exit 3 when ClawMetry has no data source
running (safe to skip silently), and support --json.
# 1. Re-read tax: files you paid to read more than once
clawmetry waste --since 6h --json
# 2. Spinning check: forward progress + loop signals
clawmetry progress --json
# 3. Spend check
clawmetry usage --json
| Signal | Action |
|---|---|
A file with reads > 3 in waste | Stop re-reading it. Write its structure into your working notes once; use offset-ranged reads for big files. |
High ratio (tokens per state-delta) in progress | You are burning tokens without producing new state. Write down what you tried, change strategy. |
loop_signals non-empty | You repeated near-identical calls. Do not retry the same thing; diagnose instead. |
Cost spike in usage | Check usage --by model: are you using a bigger model than the task needs? |
clawmetry sessions --active # who is running
clawmetry sessions <SID> --transcript # what actually happened
clawmetry sessions <SID> --cost # where the money went (incl. sub-agents)
clawmetry sessions <SID> --errors # error events only
clawmetry activity --follow --max-events 50 # live NDJSON event stream
clawmetry usage --efficiency # A-F efficiency grade
test cases passed, vs 7 / 18 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.