Nightly Build (v0.7.8-nightly.202606270713.37c158f81)
Changes since v0.7.8-nightly.202606260727.234aa97e0:
df29bfb review: preserve guided profile custom task
47c09d5 review: preserve default when editing profile skills
b536302 review: bridge legacy review config to profiles
f4c320e review: reject empty configure model values
daa98f6 review: avoid duplicate trail auth errors
e0d56ac review: preserve default profile on configure edits
d23a5cf clear turn checkpoint IDs on policy block
8cbd8d3 review: match live token sessions by planned runs
9ae8793 review: bound TUI post-run shutdown
9368a31 review: render buffered TTY output against terminal
29c8570 review: dedupe token sessions across reviewers
51c9226 trail: hydrate selected text for line findings
f172912 review: remove slot change-agent action
a32edd4 review: finalize single-agent TUI output
c6529b2 enforce local policy after sync failure
8a30eb4 align checkpoint policy push target
83bbbfc enforce checkpoint policy during writes
854115e review: unit-test pure helpers for the review/trail commands
aee626e review: preserve line anchors for invalid trail ranges
5d5990b Preserve anchors for invalid review ranges
295232f review: collapse profile key sorting helpers
be0c7df merge: fix lint after review updates
94d534b Normalize single-line review ranges
8ec25e3 Deduplicate review helper logic
3a10a57 Suppress duplicate timeout run errors
26e3e56 review: classify start-phase timeouts consistently
b1bbb69 Classify review start timeouts
a37bcae docs: minimize entire review documentation
8f29d4d docs: remove review judges handoff
7c9aeab merge: fix lint after latest main integration
496ee0a review: pin review sessions to a single checkpoint in PostCommit
3055e84 Improve review failure and trail posting handling
1dde58e inspect: mark reviewer timeout contexts with cause
ae376a4 inspect: document canceled wait errors are not timeouts
9cfadf9 inspect: make RunMulti fan-in capacity explicit
b318a52 inspect: clarify model component boundary matching
124446b inspect: rename the command and terminology to "review"
fa22e12 inspect: reserve fan-in capacity for start-failure terminals
48d27b1 inspect: remove dead code and collapse the synthesis path
b8fa833 inspect: route RunMulti start failures through terminal markers
bdce33a inspect: require earlier agent deadline for timeout fallback
15d3f21 inspect: pin --local as interactive scope preselection
0305f64 docs: revert CLAUDE review command section
6c7a25f inspect: restore model selection when adding inspectors
1a10fdc inspect: tighten review prompts to no-slop findings
956ae3b inspect: accept runner-style structured trail findings
9a77f5b inspect: treat equal parent and agent deadlines as timeout fallback
0983a81 inspect: flush buffered TUI output in the finalizer
531792a inspect: split trail output into individual findings
8b8c615 inspect: avoid timeout fallback false positives
3e4d73d inspect: restore model selection in the advanced picker
3440616 inspect: detect timeouts from inspector context fallback
e8ccafe inspect: harden model component window bounds
79f2992 inspect: document immediate cleanup on start errors
e9ddf6d inspect: test explicit empty model session matching
215ac8a inspect: show the final judge phase in the TUI
1ecb920 inspect: auto-exit the TUI when the run finishes (no Enter to start the judge)
7f0efdc inspect: address review verdict (judge alias, prompt injection, token attribution, picker model, scratch files)
9e92e76 review: document the suffix-match tradeoff in modelComponentsMatch
25c3a6f review: single-writer RunMulti state; support disabling the inspector timeout
f55c55a review/settings: clarify model offset-matching; mergeReviewProfiles always returns non-nil
2449c74 review: clarify two-pass matched-skip and tighten modelComponentsMatch loop bound
b6a149c review: match explicit-model inspectors before default ones; clarify timedOut read safety
cefc02f review: classify inspector timeout from the wait error, not a late ctx re-sample
6c85bc9 fix(inspect): set whole_change granularity when posting a verdict to the trail
f13f029 review: group timedOut with the other wg-published field writes
b8fa10c inspect: clearer end-of-run trail confirmation (posted+link / nothing to report)
f8b130a review: make per-inspector timeout detection race-free; clarify matcher/defer
99a3148 review: simplify session-disambiguation test assertion (staticcheck QF1001)
13e6273 review: test same-model/different-thinking inspectors link to distinct sessions
fb0f37f inspect: include a trail link in the post-to-trail confirmation
478379b inspect: add a per-inspector timeout (default 10m)
2522f90 inspect: drop em dashes from config UI text
66ed712 inspect: drop per-inspector model selection from guided setup
7e186be settings: make mergeReviewProfiles pure (no in-place mutation)
ae3846e review: RunMulti falls back to cfg.Model like Run
d85a3c8 inspect: store profiles in project/local settings, user picks where
9d2b78f review: only bridge model component spans across a numeric version boundary
6d457bc inspect: validate --set-judge agent can write a verdict
cc48964 review: match session models on component boundaries, not substrings
3e20abd inspect: add output destination (local or post to trail finding)
837549d inspect: resolve the judge once and pass it into runMultiAgentPath
4b0238a inspect: align judge prompt with single-consolidating-judge model
7be889c inspect: replace judge panel + chair with a single consolidating judge
14ba59f inspect: make default judge output minimal
1689b25 inspect: stop default judge prompts from padding reports
0e7fb44 test(integration): name profile for bare review invocations
2497dd0 inspect: make model the primary slot edit, agent swap explicit
f81bdf8 inspect: drop obsolete labs/session-tracking text from --help
d93bcfb Add inspect + judge-panel handoff doc
007228b Inspect cleanup: drop dead --fix manifest lookup, fix IsZero for judges
ead8c12 Address review: refresh docs + drop legacy master wording
1fa7124 Guided judge-panel picker + inspector/judge terminology
2595b9e Add a judge panel to inspect (engine + scripted config)
3f9f039 Drop the scout alias (it never shipped outside this branch)
ffef720 Rename the command scout → inspect
7ebae08 Bare entire scout requires a profile selection (no silent default)
e4afce4 Rename the focus step and add a Custom task option
51bd6a0 Guided setup edits the existing profile instead of starting fresh
23f389e Add 'entire scout --list' to list configured profiles
39ef305 Show the review-type picker in scout --configure
34a4b83 Remove Pi-specifics from review-profiles (belong on PR #1313)
6471b1b Live model enumeration for Pi via pi --list-models
724f579 Drop fabricated "example" models for codex/gemini
9a30e1d Surface codex error envelopes instead of bare 'exit status 1'
9eae501 Trim guided-flow banners; rely on form titles/descriptions
a9ea5bf Stage 3: scout guided flow — single-screen crew + standalone master step
77e4c8d Stage 2: introduce scout command (review kept as hidden alias)
6f9da77 Support a standalone master (judge) separate from worker slots
ef60a7a Make crew picker linear: first slot, then "add another"
1481e83 Make review crew slot-based (arbitrary workers, incl. duplicate agent+model)
7382fdb Combine agent + model selection into one review-crew screen
1d8ceda Add Pi review-runner adapter; drop entire review attach
b075930 Offer a model selection in the guided review setup
9317ff6 Add entire review --agents to list profile workers
34b0ac8 Add ModelLister capability + entire review --models
c145967 Don't print configure catalog before the interactive wizard
949d1cb Add scriptable review config + agent discovery
9be5923 Remove entire review --fix/--all; add --model for single-worker runs
a1297da Make review model matching ignore provider prefix and thinking suffix
b29ad76 Address review feedback on review profiles
bc13a81 Clarify review runner adapter terminology
9165231 Support review worker models
3f4c8f8 Add review configure mode
1bff4d6 Add simple guided review setup
e3be792 Use opinionated default review profile
da58e18 Add review profiles with master report