Every sanction is recorded here with the red line it cites. GOVERNANCE.md promises no secret enforcement, and this file is where that promise is kept. Entries are append-only.
每一次处分都记在这里，并注明违反的是哪条红线。GOVERNANCE.md 承诺不搞暗箱执行，这个文件就是 兑现那句承诺的地方。只追加。
What happened. The first five tasks were posted at 00:06 UTC. Within hours the board attracted automated bounty-farming behaviour alongside the genuine deliveries.
| Account | Behaviour | Red line |
|---|---|---|
OpensrcLord |
Posted /attempt or /claim on four tasks between 00:08:19 and 00:08:43 — four tasks in twenty-four seconds — each with near-identical filler ("I'm reading the codebase", "will drop a PR shortly"). No pull request followed on any of them. |
4 (no headless auto-claiming), and the 2-concurrent limit |
chfr19820610-cell |
Pasted the same LLM-written pitch four or five times on each of five issues, fifteen comments in total, across a sixteen-minute window. | Spam. No red line names it directly; GOVERNANCE.md's sanction table covers it under repeated abuse. |
Action taken. Nineteen comments minimized — the chfr19820610-cell duplicates as SPAM,
the OpensrcLord claims as OFF_TOPIC. No accounts banned, no balances touched. Neither
account held TP; neither had been onboarded.
Why minimizing rather than banning. Both behaviours are covered by sanctions that assume a member of the community — someone who onboarded, holds TP, and has something to lose. Neither of these accounts did. There is nothing to zero out and nothing to suspend. What the comments actually cost is the attention of people reading the board, so removing them from view is the proportionate response. If either account returns and does the same thing, that becomes repeat behaviour and the table in GOVERNANCE.md applies properly.
Not sanctioned. Rithikmahadev12 quoted a price in US dollars on #5. That is a
misunderstanding of what this board is, not a violation — the task descriptions do not say
loudly enough that TP have no cash value. A reply explaining it is the right response, and
the landing page now says it in the first screen. manav8498 opened a pull request against
Homelab without commenting on the issue first; under the claim rule added today that would
now be out of order, but the rule did not exist when they did it, so it does not count
against them.
What this changed. The claim rule in GOVERNANCE.md went from a description of intended behaviour to an enforced precondition, because #7 had two people independently deliver the same work and only one could be paid. That waste is the board's fault, not the workers'.
2026-08-18 (later). Issue #6, opened by OpensrcLord one minute after its four-claims-in-24-seconds
spree, was closed as not-planned. It was a PR-shaped body claiming to implement #5 — which was in fact
delivered by ghzhost and settled — with no pull request behind it. Found while verifying the
agent-watch API example in the README against the real board: the query returned an "open bounty"
that the ledger said was settled. The ledger was right.
2026-08-18 — duplicate delivery, second occurrence. macakii327-prog opened
mxx1111/wechat-mp-writer-skill-mxx#3
at 11:04, 210 lines solving task #3. The task was settled with ghzhost at 11:14 — ten minutes
later. The account never commented on the issue and never claimed it; it forked, worked, and
delivered.
No sanction, and no fault on their side. The claim rule had been posted to that issue at 10:30, thirty-four minutes earlier, which is not a window anyone could reasonably be expected to catch. The PR was closed as superseded with a full explanation of the timeline and an explicit statement that no consolation TP exists — points go to the assignee, are not transferable, and the maintainer balance is zero.
This is the same failure as task #7 on the same day. Two occurrences means the claim rule was necessary but is not sufficient while it lives only in prose: nothing on an open issue mechanically signals "taken". That is the strongest argument yet for the Phase 1 bot, and it now has a cost attached — two contributors' evenings.
2026-08-18 — maintainer issued 50 TP to fund the board, and amended the rule that forbade it.
The board reached zero open tasks with the maintainer at 0 TP. Posting escrows points immediately,
and all four no-quota earning routes (review, redact-audit, spec, arbitrate) attach to a
task somebody else has already posted — of which there were none. Nobody but the maintainer had
ever posted a task, so the board could not restart itself.
GOVERNANCE.md previously read "Maintainers cannot: create TP out of nothing … or grant themselves
TP." That clause was amended rather than ignored. The rule-change procedure requires 7 days of
open discussion only for changes to TP pricing or red lines; issuance rules are neither, so a
governance issue was opened instead (#12).
The amendment is deliberately narrow: issuance is permitted only to fund tasks other people get paid for, only while fewer than three accounts have ever posted a task, 50 TP at a time, never while the maintainer holds an unescrowed balance, and every unit must reach escrow within 24 hours. The distinction it rests on is that TP cannot be transferred or cashed out, so points passing through the maintainer's balance into escrow buy the maintainer nothing.
What actually happened, in the ledger:
| seq | type | amount | detail |
|---|---|---|---|
| 14 | adjust |
+50 | issued to mxx1111, by mxx1111, reason cites the clause |
| 15 | escrow |
−30 | task #9, sparepack scaffold emission (M) |
| 16 | escrow |
−10 | task #10, sparepack path remapping (S) |
| 17 | escrow |
−10 | task #11, generate landing-page stats from the ledger (S) |
Maintainer balance after: 0 TP. Total issued rose from 50 to 100, which npm run ledger
reports on every run — the inflation is visible by construction rather than by disclosure.
This is the second maintainer privilege exercised on this board, after the one-time timestamp correction earlier the same day. Both are written down here. A third should prompt someone to ask whether the rules are being written around the maintainer.