Publish a code of conduct and the issue templates that ask for evidence

[iderex]

A public repository is expected to carry a small set of files, and GitHub reports
which of them are there. One of seven is:

gh api repos/Flowfin/core/community/profile --jq '.files | to_entries[] | "\(.key)\t\(if .value == null then "missing" else "present" end)"'
code_of_conduct	missing
code_of_conduct_file	missing
contributing	missing
issue_template	missing
license	missing
pull_request_template	missing
readme	present

Four of those rows already have somebody on them. The licence is entry 1 of #1
and its file is #12, the contributing document is #23, the pull-request template
is #25. The two nobody holds are the code of conduct and the issue templates, and
this issue takes both because they are the same act: telling a stranger what this
repository expects before they spend an evening on it.

What this produces

A code of conduct at the root, with a reporting address that reaches a person and
a sentence saying what happens after a report arrives. An address is the part
that matters. A code of conduct with no route to a human is a document that
becomes useless at exactly the moment somebody needs it. Which address, and
whether it is the same one #24 publishes for a vulnerability, is entry 5 of #1
and is not decided here.

Issue templates that ask for the three things an issue on this board says: what
is wrong, what the evidence is, and what done means. Where the evidence is a
number, the template asks for the command that produced it, which is the request
#25 makes of a pull request and for the same reason.

A line in each template saying that nothing reads it. A filled-in form is not a
check, and a template that looks like enforcement is worse than one that admits
it is a prompt.

What this deliberately leaves out, with the reason

No CODEOWNERS. It routes review to an owner, and this repository has a single
maintainer, so the file would name one account and add nothing the ruleset on
main already requires. It also goes wrong silently on the day a second person
arrives, because nothing refuses a stale owner.

That there is one maintainer is written here as a claim, because the command
that lists them needs push access to this repository and is therefore not one a
reader of this board can run.

No governance document. How decisions are made here is #1 for the ones only I can make and the records in #2 for the rest. A third file restating
both is a copy that drifts against them, and the drift is invisible until
somebody acts on the stale copy.

Done when

The command above reports code_of_conduct and issue_template as present, the
code of conduct names a reporting address, and every template asks for the
problem, the evidence, the command behind any number, and what done means.

[iderex]

One row of the seven quoted in the body answers differently now. The command
reads the repository rather than a tree, so it carries no commit:

gh api repos/Flowfin/core/community/profile --jq '.files | to_entries[] | "\(.key)\t\(if .value == null then "missing" else "present" end)"'
code_of_conduct	missing
code_of_conduct_file	missing
contributing	missing
issue_template	missing
license	missing
pull_request_template	present
readme	present

pull_request_template reads as present, which is #25 landing
.github/pull_request_template.md. The body already names #25 as the holder of
that row, so nothing here changes what this issue covers. It is worth recording
because the same command is this issue's own done condition, and the reading in
the body is the one somebody would otherwise compare their run against.

The two rows this issue does hold have not moved, and nothing in the tree stands
behind either of them:

git ls-files | grep -Ec 'CODE_OF_CONDUCT|ISSUE_TEMPLATE'
0

[iderex]

The templates landed as .github/ISSUE_TEMPLATE/problem.md and
.github/ISSUE_TEMPLATE/decision.md, with .github/ISSUE_TEMPLATE/config.yml
beside them, on f8ee610. GitHub reads both of
them:

gh api graphql -f query='{ repository(owner:"Flowfin", name:"core") { issueTemplates { name filename } } }' --jq '.data.repository.issueTemplates[] | .filename + "\t" + .name'
decision.md	Something here has to be decided
problem.md	Something here is wrong

The command in the done condition above did not move with them, and it cannot:

gh api repos/Flowfin/core/community/profile --jq '.files | to_entries[] | "\(.key)\t\(if .value == null then "missing" else "present" end)"'
code_of_conduct	missing
code_of_conduct_file	missing
contributing	missing
issue_template	missing
license	missing
pull_request_template	present
readme	present

That field reports one path, .github/ISSUE_TEMPLATE.md, and a directory of
templates never satisfies it. Public repositories answer it in both directions:

for r in Flowfin/core cli/cli rust-lang/rust axios/axios psf/requests; do printf "%-16s %s\n" "$r" "$(gh api repos/$r/community/profile --jq 'if .files.issue_template == null then "missing" else "present" end')"; done
Flowfin/core     missing
cli/cli          missing
rust-lang/rust   missing
axios/axios      present
psf/requests     present

The three reading missing each carry a directory of templates and no single
file. The two reading present each carry the single file. So the row tracks the
older shape rather than whether a repository has issue templates at all, and the
condition as written asks this board for a file that would sit beside the
mechanism the chooser actually reads, exist only to move a field in a report, and
go stale the first time somebody edits the templates and not it.

What the condition was reaching for is answered by the GraphQL field in the first
block, which reads the templates GitHub parsed rather than a path. That is the
command worth comparing a run against, and it is why this half is finished even
though the row it was written against still reads missing.

The code_of_conduct row is untouched and nothing in the tree stands behind it:

git ls-files | grep -Ec 'CODE_OF_CONDUCT'
0

That row is where this issue stays open, for the reason the body already gives
rather than a new one.

[iderex]

Reopened. The close was mechanical rather than a judgement that the condition was
met: the message on f8ee610 carries the words
"does not close #107", and GitHub reads close #107 out of that sentence without
the negation in front of it.

Nothing about the state of this issue changed at that commit beyond what the
comment above records. The code of conduct is still absent and this is still open
on it.

[iderex]

A second row of the seven this issue's condition reads has moved, and it is not
one this issue holds. Read at 74d0531e7c4d63352733169456c554bd072daca4.

gh api repos/Flowfin/core/community/profile --jq '.files | to_entries[] | "\(.key)\t\(if .value == null then "missing" else "present" end)"'
code_of_conduct	missing
code_of_conduct_file	missing
contributing	missing
issue_template	missing
license	present
pull_request_template	present
readme	present

license reads present. The body places that row with entry 1 of #1 and with
#12, and it arrived through neither of them:

git log -1 --format=%H origin/main -- LICENSE
3ca26e9e6b6fbab522245d94fe329065182bd611
gh issue view 157 --json number,state,stateReason --jq '"\(.number) \(.state) \(.stateReason)"'
157 CLOSED COMPLETED
gh issue view 12 --json number,state --jq '"\(.number) \(.state)"'
12 OPEN

I am recording it for the reason the first reading of this command was recorded
here, which is that the same command is this issue's own done condition and the
paste in the body is what somebody compares their run against. Nothing here
changes what this issue covers, and I have changed no condition and no scope.