Tracker: `.agents/skills/backlog-burndown/SKILL.md`

[ptr727]

Tracker for .agents/skills/backlog-burndown/SKILL.md. Filed under #1315's roll-up of the open prose backlog: every finding a bundle issue enumerated against this unit is listed here once, one line each, naming the issue it came from, and the bundle is closed as a duplicate of this tracker so the finding is not lost and the bundle stops being a unit of work. A new pre-existing finding on this unit gathers here per pr-review-conduct's Merge Gate, numbering continuing this list. This tracker closes when the unit is rewritten and each line below is settled or superseded.

1. 'prose-heavy' and 'bloats' are the section's own trigger and are never defined, and a mixed code-and-doc group has no answer. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 44)
2. 'unit' is load-bearing and undefined, and not every issue names units, leaving the first bullet either vacuous or unbounded. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 45)
3. 'The change stays inside the units the issue names' and 'Deleting a claim beats qualifying it' give opposite instructions with no precedence stated. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 46)
4. 'remove the rule that leaned on it' has no test for what survives, so disproving a rationale deletes the obligation it justified. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 47)
5. The deletion trigger requires no verification that the review's disproof is accepted, so a mistaken reviewer finding deletes a live rule. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 48)
6. 'is the next issue, filed' has no actor and no timing. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 49)
7. The lead-in says 'its own bound' singular against three bullets, only one of which is marked as brief content. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 50)
8. Nothing in the section is prose-specific except its heading, so a code group's worker silently gets no review-pass budget. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 51)
9. The rule does not distinguish a neighbouring section that merely reads inconsistently from one the change has put in contradiction, and defers both. [.agents/skills/backlog-burndown/SKILL.md > Bounding a Prose Group] (Pre-existing defects in the five review-loop canonical units, found by the #1312 passes #1327 item 52)

[ptr727]

Eleven further pre-existing findings from the whole-unit reads the #1317 class 3 pull request (#1383) ran over three units of this file, gathered here per the tracker's rule. Each read verified that the references the class replaced in that unit resolve to a heading that exists and carry what the step number did.

10. The heading says "Invoking" while the grant bullet hinges on "naming" the skill, so a reader cannot tell whether an invocation through the trigger description, with no name typed, carries the grant or only the literal name does. [.agents/skills/backlog-burndown/SKILL.md > What Invoking This Skill Authorizes]
11. The unit lists one grant and four denials and leaves every other outward write the run makes, opening feature and promotion pull requests, filing deferral issues, and posting claim comments, in neither list, so under the confirm-before-outward-writes default a reader must ask before each, reintroducing the per-boundary idling the grant exists to prevent. [.agents/skills/backlog-burndown/SKILL.md > What Invoking This Skill Authorizes]
12. "A per-round merge question would idle every agent at every boundary, which is the thing this skill exists to avoid" names a purpose "Why This Exists" does not list. [.agents/skills/backlog-burndown/SKILL.md > What Invoking This Skill Authorizes]
13. The intro requires reading direct-push commit messages for their issues, while "Assembling the Promotion Body" reads only pull request numbers and their explicit line and forbids sweeping mentions, so a direct push's issues have no rule that derives them. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
14. "Open it whenever develop is ahead of main" treats the rev-list count as sufficient: an already-open develop -> main pull request makes the open fail with nothing said about driving the existing one, and a range holding only a back-merge of main gives a nonzero count with an empty diff GitHub refuses, so "a count of zero is the only case with nothing to promote" is false. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
15. Step 5 says the orchestrator "retires that group's worktree ... and then dispatches a fresh worker" during the freeze step 6 forbids dispatching under, so the intended timing, after step 7's merge, is recoverable only by reading step 6 against step 5. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
16. "nothing in the range names an issue directly" is false where the squash-merge default message carries the pull request body, since the Closes on promotion: line then lands in the range commit. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
17. "A worker also cannot finish the procedure from inside its own worktree, since removing that worktree leaves it with no working directory in which to delete its branch" is false as a tool claim: git worktree remove of the current worktree exits 0 and a git -C <base clone> branch -D run afterwards succeeds, tested on git 2.47.3, so the shared-checkout reason is the only one that stands. [.agents/skills/backlog-burndown/SKILL.md > Dispatching a Worker]
18. The brief never requires the worker to report its checkout path or a standalone-clone fallback, so "that group stops for the maintainer with the clone named" and the claim that cleanup covers every finished group are not achievable from the brief, since a standalone clone is registered nowhere. [.agents/skills/backlog-burndown/SKILL.md > Dispatching a Worker]
19. The cannot-dispatch disposition ("its worktree is then retired ... the group re-dispatched") names no actor and no moment, and does not say whether the standing branch keeps its claim comment while it waits. [.agents/skills/backlog-burndown/SKILL.md > Dispatching a Worker]
20. "which is the finished ones and the abandoned ones" is stated as exhaustive, but the local-strict-review bullet above it defines a third case, a clean tree retired with its branch left standing. [.agents/skills/backlog-burndown/SKILL.md > Dispatching a Worker]

[ptr727]

Three further pre-existing findings from the granted-round read of "The Promotion Boundary" on #1383, gathered here per the tracker's rule. That round changed only step 2's end state, which the read did not fault.

21. Step 5's retirement of the parked holder's worktree is stated unconditionally, but the "Cleanup Is the Orchestrator's" disposition it cites refuses a dirty tree, the ordinary state of a stopped or parked holder, and the unit gives no path when retirement is refused: whether the promotion fix still takes the file, or "a promotion fix outranks any file claim" is itself suspended. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
22. "the one "Grouping and File Claims" requires every feature pull request to carry and every worker brief to ask for" is false for its second half, since the brief requirement sits under "Dispatching a Worker". [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
23. "so one reported in a closing paragraph and never actually asked about stalls every round behind it": "one" binds by guess, the nearest singular nouns being "merge" and "clock" where the referent is the ready pull request from the step's first sentence. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]

[ptr727]

One further pre-existing finding from the diff pass on #1383's granted-round commit, gathered here per the tracker's rule:

24. The section's opening sentence, "driven to green and left for the maintainer", states the promotion loop's end state as green, which pr-review-conduct "Expected review loop" defines as the state before its Merge Gate, where step 2 of the same section and drive-pr step 8 both state it as every Merge Gate item except the maintainer's explicit permission, so an orchestrator reading the intro puts a pull request to the maintainer without the gate's unreadable-shape item. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]

[ptr727]

Three further pre-existing findings from the pronoun-round read of "The Promotion Boundary" on #1383, gathered here per the tracker's rule. That round changed only "it" to "that loop" in step 2, which the read confirmed binds.

25. Step 2's review-round budget stops the PR-hosted loop on a round count, where pr-review-conduct "Expected review loop" says a round count is not a stopping condition there and names local-strict-review "Disposing of Findings" as its only bounded counterpart, and the precedent it cites, "Bounding a Prose Group", answers an exhausted budget with land-and-file rather than stop-and-ask, so the two disciplines are not the same one. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
26. "with a review-round budget set before the first one": "one" has no well-formed head, since "review-round" appears only as a modifier and the nearest head nouns are "budget" and "loop". [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
27. Steps 6 and 7 reason about "the state that merge produced", but the promotion merge changes main and a next-round worker branches from develop, so the merge produces no state such a worker runs against. What does change develop under the freeze is a step 4 promotion fix, and the round prepared under the freeze dispatches at step 7 with no re-verification against it. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]

[ptr727]

Nineteen further pre-existing findings from the whole-unit reads the #1317 class 4 pull request ran over ten units of this file, gathered here per the tracker's rule, three of them from the granted rounds. Each read verified that the pointers that class introduced resolve to a heading that exists and attribute to it only what it says. Eight further findings from the same reads repeat items 3, 13, 15, 22, and 27 or an item above and are not listed again.

28. "It does write: it comments on issues, it drives and amends the promotion pull request, and it owns worktree and branch cleanup, which "Dispatching a Worker" states in full." "which" binds either to the last item, where it is true, or to the whole list, where it is false, since "Dispatching a Worker" says nothing about how the orchestrator drives the promotion pull request or comments on issues. [.agents/skills/backlog-burndown/SKILL.md > The Two Seats]
29. "Where the repository carries no priority label, and the hub does not" asserts a fact about the fleet label set with no pointer to where that set is declared, so a reader holding only this unit cannot check the fact the default ranking branch rests on. The claim is true at the current label set. [.agents/skills/backlog-burndown/SKILL.md > Ranking]
30. "That third read is not optional here: the worktree-only disposition retires a tree and leaves its branch standing" names a disposition the unit never defines or locates, since it is defined under "Cleanup Is the Orchestrator's" and the unit names no heading. [.agents/skills/backlog-burndown/SKILL.md > Grouping and File Claims]
31. "a round run during the freeze would defer every group it formed" uses "the freeze" with no referent in the unit, since the freeze is defined in "The Promotion Boundary" steps 4 to 7 and the unit's only pointer to that section is attached to how the promotion body is assembled. [.agents/skills/backlog-burndown/SKILL.md > Grouping and File Claims]
32. "Where neither read is available the set is unknown rather than empty" is unreachable, since every branch the preceding bullet enumerates has a ref the diff read can take, so the ask-the-maintainer fallback never fires and a pushed branch with no registered worktree here is read as holding only its committed files, its uncommitted edits counted as nothing. [.agents/skills/backlog-burndown/SKILL.md > Grouping and File Claims]
33. "The question is also written on its issue" has no single referent for a group of two or more issues, and the unit does not say whether the question goes on every issue, on the one whose finding raised it, or on a lead issue. [.agents/skills/backlog-burndown/SKILL.md > Raising a Blocked Question]
34. "the no-merge substitute below for one whose has not" leaves "whose" with no possessed noun, in the one sentence that distinguishes the two dispositions' tests. [.agents/skills/backlog-burndown/SKILL.md > Dispatching a Worker]
35. "presumes the branch is reachable from this repository, which the standalone clone repo-worktree allows as a fallback breaks" names the wrong referent, since a standalone clone is a clone of this repository and its branch is a branch of it. What the fallback breaks is reachability from the orchestrator's base clone, whose ref namespace the standalone clone does not share. [.agents/skills/backlog-burndown/SKILL.md > Dispatching a Worker]
36. "A pull request that is merged, or a branch whose work is complete, means the worker died after doing the work and the group is finished." The second disjunct infers "died" and "finished" from a git read the next paragraph says answers neither, and a branch with complete work and no merged pull request is not a finished group, so a live worker mid-loop with its fix committed has its tree routed to cleanup. "Whose work is complete" is also undefined. [.agents/skills/backlog-burndown/SKILL.md > Bounding the Wait on a Worker]
37. The unit claims its wait is bounded as AGENTS.md "Delegation" requires, and its only bound is the dispatch mechanism reporting the worker gone, with "Nothing about the worktree is acted on while the answer is 'still running', however long that is" removing any bound for a worker that stays alive and never reports, so a hung worker stalls the run with no report and no end. The claim and the mechanism both predate the class 4 pointer. [.agents/skills/backlog-burndown/SKILL.md > Bounding the Wait on a Worker]
38. "so that check is what separates the two": "the two" sits between the bolded "A clean one" and "A dirty one", which the no-unmerged-commit check does not separate, and the intended pair, a clean tree with nothing unmerged against a clean tree over unmerged commits, is never named as a pair. [.agents/skills/backlog-burndown/SKILL.md > Bounding the Wait on a Worker]
39. "A holder that already pushed and has an open pull request yields by having that pull request wait ... The orchestrator retires that group's worktree" retires the tree of a group whose worker may still be live, where "Cleanup Is the Orchestrator's" permits retirement only while no worker is live in the tree and says a live worker keeps its worktree until the next round's cleanup step. Distinct from item 21, which is the dirty-tree refusal. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
40. "the two hops still read them" and "the range rather than this round is still what covers earlier work" use definite references to things the unit introduces only later, in step 1 and in "Assembling the Promotion Body". [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
41. "so that one carries a single round rather than accumulating several" is contradicted by step 1, which owes the pull request for earlier rounds' work no promotion has carried, and by the body assembly, which derives the set from the range for the same reason. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
42. "A committed tracker is a file every round rewrites, which is the contention the file-claim rule exists to prevent" misattributes the rule, which prevents two live groups holding one file at once, where a per-round rewrite is sequential. The contention a committed tracker creates is inside a round, every group's pull request editing the same file to record its outcome. [.agents/skills/backlog-burndown/SKILL.md > Run State]
43. "That pass is dispatched as a worker like any other" (step 4) and "A fix landing during the freeze adds its issue to a body already written" both presume a promotion pull request finding has an issue, where "Grouping and File Claims" records a worker's claim on the issue before dispatch and "Dispatching a Worker" briefs it with a Closes on promotion: list, and a finding on the promotion pull request has no issue unless one is filed, which the unit never says to do. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
44. "Exclude the open promotion pull request from that enumeration." binds "that enumeration" to the sentence before it, "Those three enumerate the branches", a list a pull request cannot be excluded from, where the enumeration meant is the gh pr diff --name-only scan of every open feature pull request five sentences earlier. [.agents/skills/backlog-burndown/SKILL.md > Grouping and File Claims]
45. "That pass is dispatched as a worker like any other, which is the one push the freeze permits" is the unit's first use of "the freeze", with a definite article and no definition, and steps 5 to 7 then use it as a named state, while "The Round" step 8 and "Grouping and File Claims" route the reader here for its definition. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]
46. Step 6 has the next round rank and group under the freeze, and an issue whose fix is already on develop stays open until the promotion merges, so that ranking includes issues already fixed unless "Ranking" excludes an issue a merged pull request's Closes on promotion: line names, which it does not. [.agents/skills/backlog-burndown/SKILL.md > The Promotion Boundary]