skills

10 Commits

Author	SHA1	Message	Date
SikongJueluo	11de01665a	fix(review-with-me): make review rounds adaptive	2026-08-17 19:54:15 +08:00
SikongJueluo	bfe3ede15d	docs: add context notes on skill eval conventions - document evals and trigger-set coverage for teaching and arch-code-review - note review-with-me and teach-me lack evals due to disable-model-invocation - outline minimal eval cases for the review-with-me output protocol	2026-08-17 19:54:15 +08:00
SikongJueluo	14a472ebb6	fix(teaching): prioritize instruction over probing	2026-08-16 22:36:05 +08:00
SikongJueluo	324b61200f	docs: introduce repository skills	2026-08-16 17:09:14 +08:00
SikongJueluo	0627aa1841	feat(teaching): add collaborative teaching skill	2026-08-16 16:36:13 +08:00
SikongJueluo	e565478eef	refactor(review-with-me): center reviews on code slices	2026-08-12 20:41:00 +08:00
SikongJueluo	516f00c21c	feat(review-with-me): add human-first review skill - define low-noise human decision frontiers - adapt guidance to reviewer knowledge and familiarity - support precise Git and Jujutsu review scopes	2026-08-12 15:19:22 +08:00
SikongJueluo	90e12654da	feat(jj-desc): add jj description skill	2026-07-26 21:10:58 +08:00
SikongJueluo	232f3db464	refactor(arch-code-review): streamline skill and add trigger eval set Cut rigid scoring (deduction points, fixed weights, renormalization), the mandatory 7-section report format, and numeric issue caps that burdened modern agents without improving judgment. Kept the core: minimum closed loop, principal contradiction, docs > code, facts > claims, stop rules, severity-based merge gate. SKILL.md 266 -> 92 lines (~50% smaller); evals.json drops references to removed constraints; evals/trigger-set.json adds a derived {query, should_trigger} set for run_eval.py (5 positive + 5 near-miss).	2026-07-15 23:04:53 +08:00
SikongJueluo	7ad9b69811	feat(arch-code-review): add architecture code review skill and evaluation suite	2026-04-27 19:03:37 +08:00