A cognitive architecture for the entire stack of robotics
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Decide where a newly described piece of work belongs - folded into an unlanded item, as a new item in an existing plan, as a new plan of its own, or tracked nowhere - by running the mechanical scope check against live branch and pull request state, then propose the outcome via plan mode without writing any code.…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Review the current branch against the cram2 upstream main, checking for bugs and full adherence to AGENTS.md, then produce a plan-mode plan to fix every finding (including adding missing tests). Use when the user asks for a "local code review", to "review my branch before pushing", or to "review against cram2 main".
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Create (or migrate an existing bespoke roadmap doc into) a new multi-PR/multi-session plan under .claude/personal/plans/ /plan.yaml on the personal-notes branch, cross-checked against live GitHub, then bootstrap and publish it. Invoke as "/plan-create ". Use when the user asks to start tracking something as a plan…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Publish a live status dashboard Artifact for a multi-PR/multi-session initiative tracked under .claude/personal/plans/ /plan.yaml on the personal-notes branch, cross-checked against live GitHub PR/CI/review state. Invoke as "/plan-dashboard " for one plan, or "/plan-dashboard" with no argument to publish the master…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Gather everything available about one tracked plan item (its plan.yaml entry, roadmap.md history/design context, its dependency chain's live GitHub state, and patterns from already-landed sibling items in the same track), then start it in whichever execution mode is in force - presenting a plan for approval…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Change how /plan-item-kickoff and /plan-item-resolve start work on a tracked item - implementing on their own (auto), stopping for plan approval first (plan), or asking each time (ask) - by pinning the choice in the personal-notes branch so it holds for every future session. Invoke as "/plan-item-mode…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Gather everything available about one already-underway tracked plan item (its plan.yaml entry, roadmap.md history, the real state of its branch/PR - conflicts, CI, review comments, and any unresolved review threads on its upstream pull request - and any relevant discussion on its plan's tracking issue), then resolve…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
One-time setup for this repo's personal-notes tooling - the personal-notes branch, its remote/branch/path resolution, the plan-dashboard dependencies, and the CLAUDE.local.md the SessionStart hook writes. Invoke as "/setup-personal-notes". Use when someone is setting up this repo for Claude Code for the first time…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Run one maintenance pass over a stacked-PR fork-staging workflow - reparent any pull request whose base has landed, restack branches whose parent moved, and promote every approved unblocked branch to the upstream review queue. Invoke as "/stacked-pr-maintenance [fork= ] [upstream= ] [--non-interactive]". Use when…
cram2/cognitive_robot_abstract_machine
Skill Claude Code
Report the review threads a fork branch's upstream pull request has collected, including which are still unresolved, by dispatching the fork's own upstream-reviews Action and reading its job log. Invoke as "/upstream-reviews [ | ]". Use when the user asks what the upstream reviewers said, whether a promoted branch has…
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: