SikongJueluo
413c531b1e
feat(jj-describe): support automatic commit descriptions
...
Keep the explicit jj-desc workflow while ensuring completed changes receive descriptions readers can understand without task context.
- add a user-invoked jj-desc alias
- explain the rationale for non-trivial changes
- replace temporary planning labels with concrete behavior
2026-09-26 16:44:57 +08:00
SikongJueluo
0c5dfbec9b
fix(jj-describe): make description updates reliable
...
- serialize jj commands to prevent concurrent rewrites
- use stdin heredocs for multiline descriptions
- add a two-target few-shot and fail-stop verification
2026-08-29 17:51:39 +08:00
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
SikongJueluo
4078dc551c
Initial commit
2026-04-27 17:22:39 +08:00
SikongJueluo
SikongJueluo