Reusable skills for LibreSign and other free software projects. Each skill guides a task while preserving the target project's contribution and authorization boundaries. The catalog is intentionally bounded by semantic need rather than a numeric target: new workflows should extend or compose existing skills whenever that avoids overlapping activation.
| Skill | What it does |
|---|---|
| Review a pull request | Investigates changes, tests and risks; prepares file and line-specific comments for a human to review. It does not publish a review or modify the PR on its own. |
| Create issues and epics | Checks current work, drafts actionable issues and epics, and plans genuine parent and blocking links for human review before publication. |
| Execute an issue | Takes already-scoped work through investigation, focused implementation, validation, self-review and pull-request handoff; it may coordinate temporary independent subagents when that reduces work. |
| Diagnose a CI failure | Finds the causal CI failure, distinguishes regressions from upstream/environment problems and proves a focused fix with evidence. |
| Audit repository health | Finds actionable maintenance debt, separates signal from noise, checks duplicate work and produces a report for maintainer review without publishing changes. |
| Analyze upstream compatibility | Filters upstream platform, dependency and toolchain changes against actual project usage and produces a focused compatibility report. |
Add this repository as a marketplace with codex plugin marketplace add LibreSign/agent-skills and install LibreSign Agent Skills in ChatGPT or Codex. Select a skill and provide a PR link, diff, issue context, failing workflow, repository-health question, or upstream compatibility question. Repository access depends on the tools connected to your environment.
To contribute a skill or adapt a workflow to another project, see CONTRIBUTING.md and the skill catalog design.