atevals
core-ability evals for atproto agent tooling (pdsx, atproto docs MCP)
Star
Fork
Watch

Commits on main

zzstoatzz.io
docs: use absolute tangled urls for repo links
committed
zzstoatzz.io
readme: drop the file map and file names
committed
zzstoatzz.io
readme: rewrite to say what atevals is, how to run it and why it is shaped this way
committed
zzstoatzz.io
evals: package the python harness as atevals and fence off the zig spike
committed
zzstoatzz.io
evals: record the core write task results
committed
zzstoatzz.io
evals: record the pdsx 0.1.11 + public docs MCP core results
committed
zzstoatzz.io
evals: record the deployed pdsx query fix in core results
committed
zzstoatzz.io
evals: add a guarded 0.1.10 baseline and record first core results
committed
zzstoatzz.io
evals: document the core-ability task runner
committed
zzstoatzz.io
evals: add the five core-ability tasks graded against the live network
committed
zzstoatzz.io
docs: reopen where the identity and provenance capabilities live
committed
zzstoatzz.io
docs: record the purpose-built identity + provenance MCP decision
committed
zzstoatzz.io
docs: propose five core atproto abilities for the evals
committed
zzstoatzz.io
evals: split docs references into required and optional points
committed
zzstoatzz.io
evals: add docs task grading answers and atproto.com citations
committed
zzstoatzz.io
Correct search routing diagnosis and document unchanged eval failure
committed
zzstoatzz.io
evals: add bounded Watson retrieval with hidden ground truth
committed
zzstoatzz.io
evals: add native Zig lifecycle runner, OTel export, and CI artifacts
committed
zzstoatzz.io
evals: port native MCP integration to Zig 0.17
committed
zzstoatzz.io
evals: verify native Zig MCP transport and span capture
committed
zzstoatzz.io
evals: add isolated PDSX CRUD runner and independent verification
committed