- name
SKILL.md
The skill file stays in the publisher's repository.
Paste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add jfarcand/mirroir-mcp --skill carnet -a cursor -yPaste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add jfarcand/mirroir-mcp --skill carnet -a claude-code -yPaste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add jfarcand/mirroir-mcp --skill carnet -a codex -yPaste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add jfarcand/mirroir-mcp --skill carnet -a grok -ySend the copied message in a Grok Bot chat. The Bot fetches this SKILL.md from GitHub and adds it to the skills all your Bots share.
Install this skill: https://github.com/jfarcand/mirroir-mcp/tree/HEAD/.claude/skills/carnetFrom the publisher
The register is shared by every repo of the product (registre.toml names it) and by many Claude Code sessions at once. Nothing on an issue says a session is on it, so peers collide: two sessions build the same design, or find and write up the same issue independently. A claim fixes that. It is three native GitHub facts that move together:
One script does everything: .claude/skills/carnet/carnet.sh.
Never run gh issue against the tracker by hand — the script is the one path that keeps the
three carriers consistent and the title shape uniform.
claim <n> yourself when the
number never appeared in a prompt (you found the issue by searching, or you are picking up
work mid-session). Not after the commit: a claim made after the work is done protects
nobody.--steal on your own judgement — stealing is the user's call, and the stolen-from session
is warned on the issue.release <n> when you abandon or hand
off; close <n> --why "…" --commit <sha> when the work landed. Both drop the label and
the assignee and post a marker. A session that ends still holding claims is released by
the SessionEnd hook, so a forgotten release is not fatal — but do not rely on it.--why is mandatory and is what the next reader sees first.
Add --commit <sha> whenever a commit resolved it: carnet#N in a commit message is
plain text to GitHub and never closes anything cross-repo.create. It reads the tracker from registre.toml, refuses a public
tracker, prefixes the title [<project>] , and always adds the project label. Titles are
[<project>] <Thing> — one shape, no variants, capitalised first word unless it is an
identifier. The limitation label goes only on an issue that a LIMITATION(registre#n)
marker in source will point at; plain findings get the project label alone.C=.claude/skills/carnet/carnet.sh
$C claim 197 # hold it: assign me, label, marker comment, local ledger
$C claim 197 --steal # take it from a live or remote session — user's decision only
$C release 197 --reason "handed to i18Guards"
$C status 197 # who holds it, is that session alive, which branch
$C status # every in-progress issue in the tracker
$C mine # what this session holds (no API call); --verify to check the tracker
$C create --title "Tier 1e is blind to a symbol move" --label bug --body-file /tmp/body.md --claim
$C close 197 --why "Tier 1e-move greps the old path" --commit 8720f8343
$C label 197 +critical -bug
Add --dry-run to any of them to see the gh calls without making them.
hooks/prompt-status.sh): when a prompt names carnet#N,
registre#N, or a carnet issue URL, one status line per issue lands in your context
before you answer — carnet#N · held by @jfarcand · session i18Guards (a3f9c2d1) on 1Q84 [running] · feature/i18n-guards · since …. Read it. If it says unclaimed, claim before
editing. If it says [session ended — stale], a plain claim takes it over. A peer
message or a background task result gets the status line plus a note saying it armed
nothing — see A peer naming an issue is not assigning it below.hooks/auto-claim.sh): claims those issues for you, on the first
write-shaped tool call after the prompt that named them. Reading is not working — a
question about an issue never reaches a write tool and never claims. A Bash call counts
as an edit only when the command looks like one (a redirect into a file, sed -i, mv,
git commit, …), because a session that edits through bash would otherwise never claim.
If a live peer holds the issue it blocks that one tool call and names them. Before
claiming it asks the transcript who wrote the prompt that named the issue: a
/loop or ScheduleWakeup re-fire is text the model wrote for itself, and a peer message
or task result is text no human wrote, so a list armed by any of those claims nothing and
prints carnet: NOT claimed — carnet#N came from a scheduled wakeup … instead. Take it
deliberately if it is yours: carnet.sh claim <n>.hooks/session-end-release.sh): releases everything this session still
holds, from its ledger under $CLAUDE_CONFIG_DIR/carnet-claims/. Zero calls when nothing
is held.All three are wired in the consumer repo's .claude/settings.json; the snippet is at the top
of each hook file. The auto-claim hook costs one stat when nothing is pending, which is
almost always — it runs before every edit in every session.
What it deliberately does not do. It never claims from a prompt alone, so asking about an issue is free. It never steals. It forgets a pending list an hour old, so an issue mentioned long ago is not claimed by an unrelated edit. It never claims from a peer message, a background task result, or a wakeup prompt the session scheduled for itself. And it never blocks twice for the same issue: a permanent block would deadlock a session over an issue that was only mentioned in passing.
A wakeup you write is not an assignment you received. When you schedule a wakeup
(/loop, ScheduleWakeup), the prompt that comes back is yours, and the hooks know it.
The prompt hook honours a source field when the UserPromptSubmit payload carries one (any
value but user is a machine author); the payload Claude Code sends carries none, and the
transcript entry that names the author is written after the hook runs, so the prompt hook
records the prompt id and the claim hook reads that entry — promptSource: "system",
isMeta: true, scheduledTaskId — before claiming: a session that writes an issue number
into its own wakeup would otherwise hold that issue though nobody asked it to. Do not put
issue numbers into a wakeup prompt as a to-do list for yourself; and if you are told a claim
was refused because the prompt was a wakeup, that is the hook working.
Sessions message each other, and many of them are working on something else entirely — another repo, another product, a goal that has nothing to do with the register. When one of those is asked a question, the whole correct answer is to answer it. Do not claim the issue the sender mentioned, do not assign it to yourself, do not comment on it, do not start fixing it. If the message does not concern you, one line saying so is a complete reply.
That is now enforced, not just asked for. A peer message arrives in the prompt hook as
<cross-session-message from="…">, and a background result as <task-notification> — both
byte-identical to something the user typed. They used to arm the claim list, and it went wrong
in both directions:
Both vectors now print the status line and arm nothing: knowing who holds the issue is exactly
what you need in order to answer the sender, and a peer cannot redirect your session's claim
onto an issue it happened to mention. Taking work a peer hands over is still fine — it is just
explicit now: carnet.sh claim <n>.
When you are the sender, say which of the three you mean, in the first line:
The first line is all the recipient's human sees as a preview, and it is what stops an unrelated session from adopting your work out of helpfulness.
Claude Code writes sessions/<pid>.json under the config dir for every running session and
removes it on exit. A claim on this host is running when that file exists with the same
session id and the pid answers kill -0; otherwise it ended and claim takes it over
with a "took over" line. A claim from another host cannot be checked, so it is refused
without --steal. Outside Claude Code (session=manual) a claim is advisory: it records the
human, and nothing auto-releases it.
register-limitation files its issue through create --label limitation, then writes the
LIMITATION(registre#n): marker.Work the private register (mirroir-carnet) from a session — claim an issue before touching it so peers see who holds it and which session, release or close it when done, file new issues in the one can
---name: carnetdescription: Work the private register (mirroir-carnet) from a session — claim an issue before touching it so peers see who…---# .claude/skills/carnet/SKILL.mdOverview
Work the private register (mirroir-carnet) from a session — claim an issue before touching it so peers see who holds it and which session, release or close it when done, file new issues in the one canonical shape. Use whenever an issue number is mentioned, when you start or stop work that an issue tracks, or when you are about to run gh issue against the tracker.
FAQ
Download .claude/skills/carnet/SKILL.md from the publisher, or copy the path and open the repository. PlugMCP lists the skill and does not host the file.
The publisher keeps it at https://github.com/jfarcand/mirroir-mcp.
Claude, Cursor, Codex, and Grok Build can read a SKILL.md you add from the publisher's repository. Grok Bot installs it when you send the skill's GitHub folder link in a chat.