Skill Claude CodeCodex
Executes PLAN-marked parallel work via cross-platform sub-agents with supervised autonomy, structured returns, and full workspace archiving. Invoke when the working repo has a cursor-agent-team/ checkout and the request matches: User invokes /workflow or /ultra; A PLAN marked Executor: workflow needs execution with…
Skill Claude CodeCodex
Executes prose plans using a mandatory Draft -> Review -> Final loop, with general and academic tiers and the shared AI workspace. Invoke when the working repo has a cursor-agent-team/ checkout and the request matches: User invokes /writer or @writer; User needs a paper, report, proposal, documentation, or other prose…
Skill Claude CodeCodex
Master routing skill for Cursor Agent Team (CAT), the single-conversation multi-role AI team framework: picks the right role mask (discuss, crew, workflow, promptengineer, spectranslator, writer), points to the authoritative protocols, and enforces the workspace/marker contracts. Autonomous toward upstream (CAT is an…
Skill Claude CodeCodex
Orchestrate the five closing steps — topic-tree round, response verify, closing note, handoff snapshot, commit — via closingprotocol.py; stop-at-first-failure with recovery hints. Invoke when: session wrap-up / closing; handoff / archive this session; topic tree round before ending; commit the workspace.
Skill Claude CodeCodex
Render compliant dispatch prompts (dispatchheader.py) and verify structured returns (verifydispatchreturn.py) with the two-strikes ledger. Invoke when: dispatching sub-agents; parallel task fan-out; checking a sub-agent's structured return.
Skill Claude CodeCodex
Machine-run the four cold-start checks — deployment form, version, ignore policy, orphan shapes — via catdoctor.py; --fleet audits multiple host repos read-only. Invoke when: cold start in a CAT host repo; check CAT health; after install/upgrade verification; fleet audit across host repos.
Skill Claude CodeCodex
All writes to shared aiworkspace state (topic tree, notes append, plan checkboxes, plans INDEX) go through the single write gateway catwrite.py — never direct Write/Edit. Invoke when: update the topic tree / add a round; append a note; check off plan items; rebuild plans INDEX; any write intent targeting aiworkspace…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: