git
claudeFinalize a feature branch (rebase onto the default branch, squash to one commit, verify, land), then wrap up once it has landed (archive the plan, drop the worktree's test database, remove the worktree and the branch).
2 skills
Decision gates: interviewing the owner for what they actually want, briefing the owner before they pick, gating which questions reach the owner, and a persona panel that answers questions, reads stories and reviews artifacts as the people they affect.
Plugin decide from marketplace silver-cc-tools in pySilver/cc-tools. Commands read from the repo's manifest — today.
briefRebuild the user's context before asking them to choose. Use when a turn ends in a judgement call only the user can make and the context needed to answer it lives in work they are not holding right now — a plan, an ADR, a session from another day. Produces a five-part brief: one concrete run that en
gateThe question gate: run it before any question that asks the user to rule reaches them, so the user sees fewer and better questions. A judge checks whether a recorded ruling already answers it, whether a probe, a test or a rule can, or whether it is cheap and reversible enough for an agent to decide;
interview-meExtracts what the user actually wants instead of what they think they should want, through a one-question-at-a-time interview until ~95% confidence about the underlying intent. Each question carries the agent's own guess, so the user reacts instead of composing from scratch. Use when an ask is under
panelRun a persona panel on a question before it reaches the user, and turn the answers into a question card: what each option looks like to the people it affects, a recommendation that says what it costs them, the strongest dissent, what would flip it, and whether it can be undone. Each persona answers
Finalize a feature branch (rebase onto the default branch, squash to one commit, verify, land), then wrap up once it has landed (archive the plan, drop the worktree's test database, remove the worktree and the branch).
2 skills
Ask the user as a page in an agterm overlay: visual options, side-by-side trade-offs, real code from disk, sliders with a live preview, several questions with a review. Needs agterm 0.34+ and Python 3.
1 skill · hooks
Plan with coarse tasks for high-effort executors, then execute each task with Codex in the adjacent agterm pane: agree the approach, implement through a subagent, review, clear. Needs agterm, its two-agent-mailbox-chat recipe, Codex and revmux.
2 skills
Review a markdown document with the user in plannotator, shown in an agterm session next to the agent's: annotate, send feedback or approve, and the agent answers in the next round. Needs agterm, plannotator and Python 3.
1 skill
Python language server (basedpyright) for Claude: diagnostics + navigation, resolved from the project venv first so it matches the type-check gate.
LSP
Python language server (pyrefly) for Claude: diagnostics + navigation, resolved from the project venv first so it matches the type-check gate. Alternative to basedpyright-lsp — enable only one.
LSP
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.