You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
1.16.0 review: small follow-ups (spietrza/taskcell#1113) (#1133) (c615234)
1.16.0 review: small follow-ups (spietrza/taskcell#1113) (#1146) (40eeb26)
CLI: each command declares its own options; help generated from them (spietrza/taskcell#1171) (#1175) (2c8634e)
CodeQL: fix or dismiss the open high alerts in base-branch.ts and copy-back.ts (spietrza/taskcell#1153) (#1155) (ee2665c)
Codex cost estimate: persisted cachedInInput, CI partial marker, attempt merge (follow-ups from the #1056 review) (spietrza/taskcell#1059) (#1139) (1ec3474)
credentials: one credential context per run (spietrza/taskcell#1149) (#1150) (768351c)
delivery: compute the delivered change's facts once (spietrza/taskcell#1148) (#1151) (65b348f)
desktop demo: refuse writes from the command policy, not a hand list (spietrza/taskcell#1169) (#1176) (08338eb)
desktop: delete the Tauri invoke transport branch (spietrza/taskcell#1111) (#1132) (6e9c88f)
docker: pin the same CLI versions in hub and runner images (spietrza/taskcell#1165) (#1167) (7dfc945)
Follow-ups left from #967 (scorecard rename, picker drives, runs search, stats calls) (spietrza/taskcell#1031) (#1145) (5dad33d)
hub skills: tests for add/remove PRs, #ref, ls-remote cost, update permissions (follow-ups from the #1035 review) (spietrza/taskcell#1058) (#1140) (e43026b)
hub tasks: store cost and archive, owner re-check gaps, accept a verify-failed result (from #1011 item 5 and the #1053 review) (spietrza/taskcell#1057) (#1143) (e8a6609)
hub: enforce command policy at dispatch, not in each handler (spietrza/taskcell#1108) (#1141) (77c9c7f)
hub: per-project flow, so models set in the Workflow editor drive Hub and factory runs (spietrza/taskcell#1126) (#1136) (d60e767)
hub: put the caller's capabilities on the wire and gate the UI on them (spietrza/taskcell#1110) (#1138) (ddb0915)
LLM proxy Codex plan: wider sanitiser, order test, one control-char regex (follow-ups from the #1060 review) (spietrza/taskcell#1063) (#1144) (69e294d)
metrics branch: cheaper stale-lock path, lock vs absent branch (follow-ups from the #1032 review) (spietrza/taskcell#1049) (#1142) (dd2925a)
model pools per stage: weighted rotation across providers, skip out-of-quota models, then fallback (spietrza/taskcell#1131) (#1154) (590b2c4)
revise-pr.ts: replace the raw NUL byte that hides the file from grep (spietrza/taskcell#1170) (#1173) (3bd1143)
runners: one agent-session module instead of six hand-built copies (spietrza/taskcell#1172) (#1178) (a9ee447)
skills: hub gets the PR result from a core module, not by parsing CLI output (spietrza/taskcell#1112) (#1134) (9eb40a7)
tasks: one dependency gate for GitHub, GitLab and Linear (spietrza/taskcell#1109) (#1137) (d6a3542)
watch: move per-source primitives to their source modules (spietrza/taskcell#1147) (#1152) (4cc4172)
Bug Fixes
move workflow token permissions to job level; an implement caller must now grant actions: read (spietrza/taskcell#1122) (0016470)
MR reviewer gets read-only tools like the PR reviewer (#1184) (b8eb285)
one review run per labelled PR so none is dropped (#1166) (7f2d4c7)
raise the review turn cap to 30 for Opus on large PRs (#1156) (10db1ca)
review-mr and research respect the usage cap (spietrza/taskcell#1168) (967ccdc)
revise: count bullets under a bold numbered finding as its detail, not as separate blocking items (spietrza/taskcell#1157) (78dd45e)