F206 Commit subjects carry a Conventional Commits type and the task id, which joins the history to the ledger

Field	Value
Status	active
Topic	Commits and concurrent sessions
Grade	R2/S3
Cases	roadkeep, polyweave, freewilly, winwright, Shio
Contrary	none
Harness facts	none

Claim

Under the practice nearly every commit subject opens with a Conventional Commits type and most name the task id they ship, in all five projects, and in Shio both appeared only with adoption; this is one convention, carried by the owner's commit tool and the skills that prescribe it, not five independent choices.

Mechanism

The commit tool writes Conventional Commits messages, and the skills tell the agent to pass the title itself, in that form. The backlog gives every task an id, and the ship step writes the id into the ledger, so a subject naming the id lets a reader go from a ledger line to its commit and back with a text search. Commits that only file tasks are made separately, so an id in a subject means the task shipped or moved in that commit.

Observations

Counts are this study's, made at the pins over git log <pin> --format=%s, as the case studies give them.

roadkeep. 1,938 of 1,946 subjects are conventional; the rest are merges and three untyped

commits. 1,655 name an RK id. The id's place in the subject changed three times, from "feat: load roadkeep.toml as the project's declared format (RK3)" [roadkeep@7068a102] to "docs(RK1115): file the pause no line at its why budget can pass" [roadkeep@5345a5bb], to "feat: RK1298 - one budget and its deltas, not three tables" [roadkeep@8ec64f2b] and back to a suffix in "feat(list): carry each pause's age and reason in the --stale payload (RK1677)" [roadkeep@f7333f42]. Filing is its own commit: "docs: file block J, validation, with RK1690-RK1694" [roadkeep@9ecfa187].

polyweave. Every one of 132 subjects opens with a type; 108 carry a scope and 125 name a

PW id. Tasks found during work are named in the title, as in "(PW1), with PW37 filed" [polyweave@8d1c9ab].

freewilly. 425 of 444 subjects are conventional; the other 19 are pull-request merges.

325 name the task right after the type, as in "DD112 a fold does not outlive the project it was set on" [freewilly@d01fcb3]. Filing is its own commit: "DD112-DD113 file what the container follow-ups turned up" [freewilly@c44801a].

winwright. 706 of 708 subjects are conventional and 668 name a WW id. The id first sat in

the scope, as in [winwright@d678d09], and later moved to a suffix, as in "refactor(site): one C# reader for every generator (WW502), with WW507 filed" [winwright@3f5009b].

Shio. None of the 1,778 subjects before the adoption commit [shio@6bf11b754] is

conventional and none names an SH id; of the 1,668 from it to the pin, 1,361 are conventional and 1,148 name an SH id. The index prescribes the tool that "generates a Conventional-Commits message from the staged diff via the OpenAI API" [shio@821f18d74:agents.md#L241-L242].

Independence

One origin. The subject form comes from the owner's commit tool and from a title rule written in the same words in each project's skill, and the ids come from roadkeep, which governs all five backlogs. Under rules 2 and 3 of ../grading.md that is one tool and one copied practice. Shio's change at adoption shows the convention arriving with the tooling; it is not a second origin.

Counter-evidence

None found at the pins against the convention itself. Its instability is recorded: roadkeep moved the id three times and winwright once, so a search by id has to allow for more than one position. Shio's 307 non-conventional subjects after adoption were not examined.

Grade

R2: one origin, observed continuously in separate projects and periods, with an attempted change of form in roadkeep and winwright. S3: the shares are counted at the pins with a repeatable command, git log <pin> --format=%s matched with grep -cE against a type prefix and against the project's id pattern, and what it counts is the finding's own effect: subjects in the convention and subjects naming an id. The finding says what the history looks like under the practice; it is not evidence that the convention caused any benefit, which no source measures.