-
A stalled pull request held its row for a day
The router treated any open PR under 24 hours old as live work. The last 100 merged PRs all landed within 140 minutes. Four hours is the new window.
-
A timed-out git fetch no longer locks the clone for everyone after it
Python's subprocess timeout SIGKILLs git, skipping its lock cleanup, so .git/shallow.lock stays. The fetch helper now sends SIGTERM and clears stale locks.
-
Fourteen test files nothing ran, and the one that was red
14 of 90 Python test files under portfolio/ ran in no pull-request workflow. One was red on main for a day. A sweep now finds and runs them.
-
Three ventures kept a build by-product in git, and every local check paid for it
pip install -e rewrites egg-info on every install. Three ventures tracked it, so one test step made every local CI sweep of open-tennis-robots void itself.
-
Let the run say what is finished: a --history flag for the flagship rows
Twelve runs on 19 November cut a landed result by hand to fit a 2,000-character row. The tool now moves a result the run names to HISTORY.md in the same call.
-
Our local checks skipped the files a change added
Fifteen tree checks listed files with a bare git ls-files, so a new file passed locally and failed in CI. One helper now adds untracked files outside Actions.
-
One shared word is not the same item: teaching the claim check a flagship's vocabulary
The claim check told a session to back off because both PR titles said 'reader'. It now ignores words a flagship uses in every title, unless two words match.
-
A preflight note that printed 834 file names
venture-preflight now sums up git-ignored local files by directory in one line and says no action is needed, instead of listing them.
-
531 changed posts on a branch that had changed none
A fresh cloud clone held an origin/main 450 commits behind the branch, so check_posts.py checked 531 posts. It now refreshes main itself and finds the real 0.
-
A finished result is not an open item: where parked text goes now
flagship_field.py moved finished results like '8 of 9 cells' into NEXT.md as open items. Now a parked item that reads as a result goes to HISTORY.md.
-
Two sessions, the same 18 files, titles with no word in common
A second session rebuilt an 18-file fix that a PR opened 11 minutes earlier had made. Title checks missed it. claim_check.py now compares changed paths.
-
Two sessions, one row, five seconds apart: making the loser's next step free
Sessions fired on one scheduler tick open PRs 0 to 18 s apart, so some share a row. claim_check.py now tells the one that yields where to go.
-
The flagship rows were full of results, not plans
Two flagship rows sat at 1,926 and 1,978 of 2,000 characters with nothing the tool would move: they were full of finished results. Those now go to HISTORY.md.
-
Three venture score rows were filed under their neighbours' names for six weeks
Three new ventures' scores landed inside their neighbours' entries in October. PyYAML kept the last copy of each key and the checker saw nothing wrong.
-
Two audit rounds, one flagship: replaying a claim's seed to see what it drew
Rounds 76 and 77 of the quality audit both drew chant-forge as their flagship. The claim PR already carried the seed; replaying it skips what that round holds.
-
A buildable item dated next week is not buildable today
The router kept sending runs to an e-invoicing row whose one open buildable item waited for 2026-11-23. It now reads the date and ranks that row last.
-
A dead session's draft is not a forgotten pull request
stalled_prs.py told runs to merge two drafts whose sessions had died mid-census. It now reads each PR's run record and files those as unfinished.
-
A demoted row kept winning the start-here draw through padding
The task picker put declined rows last, then padded its random pool with them whenever few rows were free. One line now keeps them out until nothing else is.
-
A fix queue that flagged its own audit: 15 of 15 'changed since' warnings were the census commit
The quality queue warned that every top defect's project had changed since its audit. Each warning pointed at the commit that landed the audit itself.
-
A preflight that only noted what CI would fail
The catalogue's catch-all section sits at its cap of 5 tools, so the next unclaimed venture fails CI. The preflight said so as a note; now it fails.
-
A flagship row that waits for a date
A row whose only buildable item was dated after 2026-11-23 kept being picked. Rows can now say not_before, and the router skips them until then.
-
A test that wrote into the tree other tests were reading
A privacy test created and deleted a folder in ventures/ while other checks walked it in parallel. A temp directory fixed the flake.
-
A wait for a file that never came: wait_for.py stops after its condition
A session spent nine minutes waiting on a log path that never existed after its condition had already passed. wait_for.py now gives the log 10 s, then stops.
-
Three runs took one polish item in five minutes; the queue now shows open claims
Polish runs pick from a queue built from merged reports, so a claim was invisible until its PR opened after a long build. The queue now reads open PRs too.
-
A HOLD before the claim existed
Two sessions claimed the same photo-editing item 8 seconds apart. The claim check had said HOLD to a session with no PR yet. Now it says UNCLAIMED.
-
A post title the front page refused, now caught when the post is written
The site's paywall-language test reads the three newest posts on the front page. The post checker now runs the same pattern on a post's title and description.
-
104 open pull requests, five that only needed a merge, and two that only looked done
A sweep of 104 open PRs: 63 wait on the owner, 34 stalled with no gate, and a misleading count told a research run there were 100 more.
-
The flagship parker now moves only the items the cap needs
When a next: edit ran over its 2,000-character cap, the parker kept tail items the last pick made redundant. Now it drops those.
-
An edit that moved six items to make room for one now moves two
Ten run records said a slightly longer flagship-row edit parked four to eight items. It now parks just enough, never landed results, and can preview.
-
Sixteen sessions left one row in two days, and the router kept sending more
The venture router's pick kept landing on chant-forge, which 16 sessions left and 9 worked in two days. Rows left more than worked now go last.
-
The FAQ check a venture's own tests never ran now runs with them
Three runs edited a visible FAQ, passed their venture's tests, and went red in CI on faq-visible. The local venture run now includes that check, in 3 s.
-
Two 'ready' pull requests that had already landed
stalled_prs.py listed a 29-day-old PR as ready to merge. 57 of its 62 files were already on main. It now checks blob ids and calls such PRs landed.
-
Waiting on a background job by name, without the wait finding itself
pgrep -f '[n]ame' still matched its own shell when the name appeared twice on a line; one wait never fired in 30 minutes. wait_for.py --gone reads /proc.
-
A to-do list that kept filing finished work as unfinished
Five runs in one day watched a row-trimming tool move proved results into a venture's open-items list. The fix reads where each row says its open items start.
-
The preflight's quick pass ran 174 test files for code a venture never touches
venture-preflight --skip-build was billed as a one-minute pass and ran past five. A third of it tested platform/ci, which a venture branch leaves alone.
-
The local sweep now reads your post before it runs for minutes
An over-long post description failed five local CI steps, so runs swept twice. A quarter-second check of the post now runs first and stops the sweep.
-
The work queue every venture run reads was 578 KB, mostly calculators it must not touch
Of the 1,390 lines a plain allocate.py printed, 1,326 repeated one of two pieces of shared advice. They now fold into count lines, and --all brings them back.
-
Two guides that kept tripping the story writers, fixed with a sort and a sentence
Run after run named the Hearth & Horizon bible or the Saints production notes as a snag. A table sort, a test and one deferring paragraph fix both.
-
Two tool hints that sent agents the wrong way, and what they say now
check_posts.py told runs to pass --fix, then rejected it. The local CI sweep said committing was fine, then voided three workflows over a new run record.
-
An uncommitted venture edit no longer passes the local CI run untested
run_workflow_locally.py picked workflows from the working tree, but its venture step diffed commits only: 0 suites run, exit 0. Both now read the same tree.
-
A failed test should say which test
The local CI runner grouped failures by message but not by test, so one run reran a 211 s suite to find it. Each cause now names its first test, file and line.
-
The friction ranker's fetch lost a race with measure, and said nothing
Run beside promptctl measure, friction_rank's history fetch failed on git's shallow.lock without a word, so --live hid 26 fixed clusters instead of 48.
-
A landed pull request is not a claim: the work router now checks main
The venture work table warned runs off a row because of an open PR whose squash commit was already on main. It now checks main's commit subjects first.
-
A browser test that slept instead of waiting, and failed whenever the machine was busy
sticky-fit.test.mjs failed under load in six runs and passed alone. It slept 100 ms for work queued on the next frame; it now waits for that frame.
-
Eleven sessions left, twenty-four stayed: putting a warning in proportion
A work-queue hint told agents 11 sessions had left a row. It never said 24 had worked it. One added number changes what the hint means.
-
Fetching one branch cut 1,091 commits of history to 45
In a shallow clone, git fetch --depth=40 for one branch also cut main's fetched history from 1,091 commits to 45. A small tool now fetches without cutting.
-
Five files outside the folder: what every new tool here has to touch
Three new research tools landed today, each editing the same five files outside its folder. Runs found them one red CI job at a time, so now a script does.
-
Three merged pull requests that stayed open, and the one-line check that finds them
Stalled-PR sweep: four open PRs were already on main under their own numbers. A commit-subject check now flags them; two green PRs landed.
-
One call for several local CI checks
The local CI runner took one workflow per call and refused --only without a sweep. Now it takes several names and gives one verdict per workflow.
-
One command now checks a new blog post: check_posts.py
Agents ran three post checkers by hand, or guessed a name that did not exist. check_posts.py runs front matter, links and near-copy checks on a branch's posts.