Skip to content
v1.0.0MIT

Copilot CLI workflows that pause for your approval: evidence-backed standup drafts, review requests and review-status reactions, PR feedback handling, draft PR sweeps, careful rebases, shared workflow version updates, investigations, coordination across environments, and proposal-first improvement of the shared context.

assess-and-address

Work through every open pull request review thread and top-level review comment, judge whether each one is right, commit focused fixes for the valid ones, reply to and resolve threads opened by automated reviewers, and draft (never post) answers for human reviewers. Use when asked to address or work through PR feedback or review comments, including Copilot and other bot findings.

challenge-pass

Run an investigation's conclusions past a fresh, read-only subagent before sharing them or naming a root cause. It tests each claim for direct observation, prior history, recent change, whole-population coverage, sourcing, a more direct signal, what labels really mean, independent confirmation, fit to the original goal, and internal consistency. Use before publishing analysis, asserting a root cause, or declaring something broken, benign, fixed, safe or harmless, and whenever asked to sanity-check or challenge an analysis.

done

User-only command that wraps up this chat with a short recap, then the session-done extension saves a [DONE] marker on the session title and switches to a fresh session. Never call this skill on your own; only the user starts it with /digits:done.

draft-pr-sweep

Read-only triage of the user's open draft pull requests that sorts each by its next move (a blocking label, an unmet dependency, merge conflicts, CI state, a review already in flight, feedback to assess, or readiness for review), offers follow-up actions in one approval form, and hands approved work to the rebase-pr, assess-and-address and request-review skills. Use when asked to sweep or triage my draft PRs or report where they stand.

inter-env

Check progress across your active Digits environments, draft scoped instructions between chats, and deliver individually approved messages through existing tmux panes; use for coordinating environments, handoffs, checking on other chats, or sending instructions to another environment.

investigate

Evidence-first investigation of incidents and anomalies. Starts from the original artifact (alert, log query, chat thread, ticket, pull request, CI run, incident or data anomaly), finds the change that introduced the behavior, challenges its own conclusions, weighs options including doing nothing by impact, blast radius and cost, collects every decision in one batch, and produces only approved follow-ups. Use when asked to investigate, triage, dig into or look into something, or to explain why a failure, alert, incident or anomaly happened.

rebase-pr

Rebase one GitHub pull request onto the current tip of its base branch, with an ownership preflight, a dedicated worktree, conflict resolution that keeps what both sides intended, a semantic review of code that merged cleanly, verification, and a lease-guarded force push bound to the previously recorded head SHA. Use for conflicting or out-of-date PRs and when asked to rebase a PR, bring it up to date, or fix its merge conflicts.

request-review

Drafts and, once you approve, posts a short request asking for review of one or more PRs in your configured review channel, mentioning the reviewers you name or else the reviewers currently requested on the PR; with approval it can also take a draft PR out of draft. Use when asking for a review, nudging reviewers, making a PR ready for review, or sharing a PR in the review channel.

review-reactions

Keep status reactions current on the user's own review-request messages in the configured team chat channel - one for comments from humans other than the PR author, one when every linked PR is approved, and one when every linked PR is merged. Plans as a dry run first and adds reactions only when the user approves. Use when asked to sync, refresh or check the status reactions on review requests.

self-improve

Reviews work chats from every environment that are newer than the last completed run, proposes evidence-backed improvements to the shared context, and changes nothing until you approve the proposal. Runs only when the user explicitly asks or types /digits:self-improve, never as a side effect of normal coding or standup work.

standup-update

Builds your daily standup or progress update in your configured format by cross-checking earlier updates, recent work chats, issue-tracker data and live GitHub state. Use for any request to write, draft or generate a standup, daily update, check-in or progress update.

summarize

Read-only status recap of the current conversation (goals, progress, blocking questions, next steps) that changes nothing except saving the same recap in this session's workspace. Use for "where do things stand", "recap this chat", "status of this chat", or preparing a handoff to another chat.

update-shared-workflows

Check and bump pinned versions of an organization's shared GitHub workflow and action repositories. Finds each repository's newest tag (repositories come from arguments or bindings), inventories pinned references in one repository, reads release notes for breaking changes, then applies approved same-major updates through a preview-first helper. Use when asked to check, bump or update shared workflow or action versions.