fail
animate
It missed 3 of 39 requests it should have handled — the skill stayed silent and the model answered on its own.
Fired when it should have
fired in 36 of 39 attempts
95% confidence between 80% and 97%. 39 attempts settle it to a narrow range.
Was right when it fired
was right in 36 of 36 attempts
95% confidence between 90% and 100%. 36 attempts settle it to a narrow range.
1 observations could not be read
Every case
Instability
3 cases produced both outcomes
- trigger.positive.sidebar_collapse — 9 passed, 1 failed
- trigger.positive.hover_card_lift — 9 passed, 1 failed
- complete.writes_toast_motion — 8 passed, 2 failed
Conditions
Two runs are comparable only when all of these are identical. A score that moved because the model changed is not a regression in the skill.
- Skill version
- emilkowalski/skills@d23d7f88a2e21c9e4b1418c7abe420f5c1052ba7
- Skill hash
- sha256:45ba81da9f54aa1d53cb7a5a9cddb141222cad7b11e1c1545175479fd2577d69
- Model
- claude-haiku-4-5-20251001
- System prompt hash
- not-provided-by-host
- Environment hash
- sha256:06935ec7fdf935afec04e70a1c8c39ac1e6f8534978ce40d6cbf622ecc29e6ff
- Permission mode
- not reported by the host
- Host memory
- not measured (the record predates 0.4.5 or the probe did not run)
- Context hash
- not measured (the record predates 0.4.5 or the probe did not run)
- Case set version
- 1
- Case set hash
- sha256:789aa7faedba46d0adea3c8d1971f39c52a616eb3623ebe466f9fcb88973fbf1
- Assay version
- 0.3.1 or earlier (the record predates version stamping)
- Activation check
- not made: the record predates 0.2.0, which began confirming that a selected skill actually loaded
Recorded before the host CLAUDE.md leak was closed (Assay 0.4.5). This run’s working directory sat under the measuring machine’s home directory, and Claude Code loads instruction files from every directory above the one it works in, so that machine’s ~/.claude/CLAUDE.md — a single instruction about an unrelated tool, graphify — would have entered the context. Which instruction files actually entered it was not measured.
The name graphify appears in none of this run’s 100 attempts. No trace does not prove no effect.
Not re-measured: no Skill call in any published run targets graphify, and the one finding the file could plausibly affect — the marketing-skills positioning case, 10 of 10 attempts — was reproduced without it, 8 of 10.
What it cost
- Attempts
- 100
- Tokens in / out
- 4,759 / 204,458
- Cost
- $4.90
- Agent time
- 49 min
Combined trigger score (F1) 0.96 · 96 of 99 trigger decisions were correct.