The five full-width screenshots pushed the install command five scrolls down.
One image now, with the others as links.
The room shot is retaken on a room with real traffic: two agents disagreeing
about whether a 402 is terminal, escalating to the operator, and the operator
ruling -- with the todo sidebar open. The old one was a wall of SYSTEM lines
about worktrees, which showed the plumbing rather than the product.
CHANGELOG.md and the title's version are GENERATED by sync.sh from the source
repo, not typed here, because a hand-kept copy drifts -- which is how the
install URLs ended up pointing at a private repo. The version comes from the
newest tag by creation date, not by semver: the version line was reset flat at
0.1.0, so v0.7.0-rc1 still sorts higher than the current release.
The changelog stops at 0.1.0 (below it is the retired rc lineage, which reads
higher) and starts at 0.1.1 (0.1.2 is mid-release and not tagged yet).
Signed-off-by: disqualifier <dev@disqualifier.me>
Shot against a live 0.1.1 daemon, dark by default because that is what the
product ships: readout.css defines dark on :root and light as the
[data-theme="light"] override. One light shot of the same KB page, so the
pair shows the difference rather than two unrelated pages.
Real data rather than a fixture, so what they show is what the product
actually looks like with work in it.
Signed-off-by: disqualifier <dev@disqualifier.me>
The release rewrote the pinned tag in both files. Their header comments now
name the private source repo at v0.1.1, which users cannot fetch; the README's
own install commands point at this repo's raw/branch/main and are the ones
people actually run. Filed as a defect rather than patched here: the pin
rewriter and C24 both require a tag URL in these two files, so changing them
to point at this repo is a design decision, not a sync fix.
Signed-off-by: disqualifier <dev@disqualifier.me>
It was written when --runtime codex refused outside a project, so it told
people to run the whole thing per project. The skill and hooks are
machine-wide now; only the AGENTS.md room rules are per-project, and the
/hooks approval inside Codex is the other step the installer cannot do.
Signed-off-by: disqualifier <dev@disqualifier.me>
Synced from the source repo. install.sh and install.ps1 now run
`standup install --runtime codex` machine-wide where codex is on PATH,
with --no-codex / -NoCodex to opt out, and name the per-project
AGENTS.md step in their closing lines.
Signed-off-by: disqualifier <dev@disqualifier.me>
The installer cannot run `--runtime codex` for you: the verb requires a
project, since one of its targets is the project's own AGENTS.md. It now
detects codex and prints the command to run in each project instead.
Signed-off-by: disqualifier <dev@disqualifier.me>
install.sh / install.ps1 synced from the source repo: they now wire the Codex
runtime where codex is on PATH, with --no-codex / -NoCodex to opt out.
README: say what gets wired and name the Windows flag spellings; drop the
stiffer wording in the licence and service sections; add a table of where the
binary, the data and the service definition actually live, since ~/.standup is
the thing people need to know about for backups and for STANDUP_HOME.
Signed-off-by: disqualifier <dev@disqualifier.me>
install.sh and install.ps1, copied from the agent-standups source repo, which
stays private. Users need this repo and the package registry; they never need
the source.
The files are copies rather than symlinks on purpose: Gitea's raw/ endpoint
serves a symlink's link text, so a symlinked install.sh would pipe a relative
path into the user's shell instead of a script. sync.sh refreshes them from
the source tree and is a no-op when nothing changed.
agent-standups remains canonical for both files: the probe harness reads them
from its repo root (run-all.sh passes $REPO/install.sh to the R06 probes) and
P8-24 copies them into a scratch tree.
Signed-off-by: disqualifier <dev@disqualifier.me>
dsql