t-step
Community@t-step
t-step publishes Claude-native skills governing repository issue resolution, release decisions, license auditing, PR scoping, and session continuity for disciplined software delivery.
Agent Skills by t-step
Showing 14 vetted skills indexed across 1 GitHub repositories.
issue-work-loop
Orchestrates a six-phase workflow for taking GitHub issues from discovery to verified close-out.
Skill-Name-With-Hyphens
Provides a reusable template for authoring new Skill units with standard sections.
release-captain
Recommends release version and timing from evidence, then orchestrates gated release-PR creation.
domi-consumer
Detects .domi-pin drift status and reports inherited DomI policy categories in consumer repositories.
maintain-claude-md
Creates, updates, and lints CLAUDE.md files for Claude Code sessions.
repo-hygiene-init
Bootstrap baseline repository hygiene including pre-commit hooks, linting, formatting, Makefile, CI, and licensing.
package-release-integrity
Validates Python package release consistency across version sources, tags, changelogs, and semver movement.
session-continuity
Organizes session notes, handoffs, and project profiles in a notes home outside project repos.
license-compliance-auditor
Audit repositories for license compliance across dependencies, vendored code, fonts, assets, and snippets.
fork-pr-flow
Guides opening and targeting pull requests across fork and owned-repository Git workflows.
context-graph
Drafts and reviews semantic relationship proposals through the deterministic context-graph CLI.
hands-on-keyboard
Guides Claude to coach users through coding tasks instead of editing files directly.
verify-then-commit
Runs tests, typecheck, lint, and format checks before allowing any git commit.
scoped-sequential-prs
Splits large changes into ordered, single-purpose PRs with a mechanical contamination gate.
Frequently Asked Questions About t-step
FAQPage SchemaWhat tasks can I accomplish using t-step's skills?▼
You can drive issues from discovery to verified close-out, get evidence-based release version recommendations, audit license compliance across dependencies and vendored assets, scaffold repo hygiene baselines, maintain CLAUDE.md files, split large changes into scoped sequential PRs, and verify tests before committing.
Who are t-step's skills designed for?▼
They target software engineers and maintainers working with Claude on repository governance — particularly those managing releases, pull request hygiene, license obligations, and multi-session projects where honest status reporting and human-authorized publication gates matter.
How do the release-decision skills work in practice?▼
release-captain gathers evidence since the latest tag, recommends a version class and timing with rationale, then only on explicit per-step human approval drives a Release Please PR. It never bumps versions, edits changelogs, tags, or publishes — package-release-integrity acts as its safety check.
What prerequisites or dependencies do these skills require?▼
Skills operate inside Claude skill environments and reference existing project infrastructure: git repositories, test/typecheck/lint commands, Release Please configuration, and optionally a .domi-pin file marking DomI-inherited policy. The context-graph skill requires the deterministic context-graph CLI for endpoint legality and candidate keys.
Do these skills ever commit, merge, or publish autonomously?▼
No. The manifest repeatedly separates implementation authority from close, merge, and publish authority. verify-then-commit enforces running tests before committing, fork-pr-flow prevents self-merging, and release-captain never cuts releases — publication always requires explicit human authorization.