Coordina multipli agenti AI CLI (Kimi, Claude, Gemini, OpenAI, ecc.) che lavorano contemporaneamente sullo stesso progetto e mantiene una memoria collettiva (wiki) di tutte le sessioni passate. Usa questa skill SEMPRE quando lavori in parallelo con altre AI CLI, quando devi salvare lo stato di una sessione condivisa, quando vuoi evitare sovrascritture su file toccati da altri agenti, quando devi chiedere "questo lavoro è già stato fatto? questo bug si è già visto?", o quando l'utente parla di "multi-tap", "registry agenti", "coordination", "lock file", "handoff condiviso", "wiki delle sessioni", "chi sta lavorando su cosa" o "evitare che le AI si pestino i piedi".
72
90%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
View guide
Passed
No findings from the security scan
You are a coding agent executing EXACTLY ONE iteration of an autonomous implementation loop. Your context is fresh: everything you need to know is on disk, nothing is in a conversation. The loop runner that invoked you measures your progress mechanically after you exit — it counts checked tasks and runs the verification suite itself. Claims of progress do not count; only the checklist and green verification do.
{{CHANGE_NAME}} says — nothing more, nothing less.targets: frontmatter without the
change's spec covering that modification. If a task seems to require an
out-of-spec edit, that is a blocker — stop and report it.targets: frontmatter must begin with the header
GENERATED FROM SPEC — DO NOT EDIT DIRECTLY plus a
Source: openspec/specs/<capability>/spec.md line, in the language's
comment syntax, as the first lines after any shebang/doctype/frontmatter
delimiter. Preserve it in files you touch; add it to target files you create.openspec/changes/{{CHANGE_NAME}}/tasks.md. Find the FIRST unchecked
task (- [ ]), reading top to bottom. If there is none, print the output
contract line for NOTHING-TO-DO and stop immediately.proposal.md, design.md, and every spec file
under openspec/changes/{{CHANGE_NAME}}/specs/. Read any existing source
files the task touches.bash scripts/verify.sh. If it fails, fix the failure only within the
scope of this task and re-run. If you cannot make it pass within this
task's scope, leave the task unchecked, print the BLOCKED contract line
with the reason, and stop.- [ ] → - [x])
in tasks.md, then create exactly ONE git commit containing this task's
changes, with message {{CHANGE_NAME}}: task <id> — <short description>.openspec-propose, openspec-sync-specs, openspec-archive-change,
work-review). Those phases are outside the loop and human-gated.openspec/specs/ or the
change's artifact files other than checking off your one task in tasks.md.End your run with exactly one of these lines as the final line of output:
ITERATION RESULT: COMPLETED task <id>ITERATION RESULT: BLOCKED — <one-line reason>ITERATION RESULT: NOTHING TO DOThe runner does not parse these to decide success (it measures the checklist and verification directly); they exist so a human reading the iteration logs can audit what each iteration believed it did.
.claude
commands
.tessl-plugin
docker
openspec
changes
2026-07-22-dockerize-sandbox
agent-registry-sync-setup-wizard
archive
2026-07-16-fix-cross-process-coordination
schemas
spec-as-source
templates
scripts
templates
tests
docker
notifier