Finds open conference CFPs relevant to the user across Java/AI/developer conferences, with persistent sent/dismissed/remind state and source-aware Sessionize verification. NanoClaw per-chat overlay, loaded via containerConfig.additionalTiles.
73
92%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Low
Low-risk findings worth noting
The operator's cfp-labeling-2026-10-16.json supplies 17 non-geographic
yes/no labels: ten yes and seven no. Geo-excluded, unlabelled and
unknown_to_baruch rows were excluded. The evaluation used the current
speaker profile, relevance criteria and conference metadata, including
bounded website descriptions. No state decisions changed during calibration.
The final evaluation refreshed Sessionize metadata through the current normalizer before fetching website descriptions and recorded each full distribution. It automatically kept FOSDEM and DevFest Berlin, dismissed Dev Days Students Madrid, and routed fourteen to Sonnet. All three automatic decisions matched the operator's labels. Every response in this final pass supplied a usable distribution.
The executable routing contract and calibration accounting procedure are
referenced in skills/check-cfps/references/jev-relevance.md.
This small set selected conservative routing boundaries; it is not an independent accuracy estimate. Retain the middle band until a larger labelled evaluation supports narrowing it. Re-evaluate when the model, questions, profile or metadata projection changes.
.tessl-plugin
skills
check-cfps
references
scripts
nightly-cfp-sync