Audits and fixes repository drift against dev-env skill contracts.
npx skills add https://github.com/brandonarbini/arbini.family --skill converge-brandonarbini
Please help me install this Agent Skill. Skill: converge Source: https://github.com/brandonarbini/arbini.family/tree/main/.agents/skills/converge Command: npx skills add https://github.com/brandonarbini/arbini.family --skill converge-brandonarbini
.agents/skills/*/SKILL.md with a ## Contract section and checks each bullet as a statement that must be true of the repo right now. - Guided Remediation: Fixes deviations using the guidance in each skill's own body, while protecting dev-env-generated files from hand edits that would be silently reverted. - Verified PR Workflow: Verifies fixes inside the devcontainer with lint, typecheck, and test commands, then opens a PR titled "Converge on dev-env standards" with one commit per unrelated item. - Use Case: When CI starts failing right after shared dev-env standards change, run this Skill to audit every contract bullet, fix drift in AGENTS.md or CLAUDE.md, verify in the devcontainer, and open a reviewable PR. ## Quick Start Converge this repo on the current dev-env standards and open a PR with the fixes.High-intent search queries and answers about installing and using this skill.
Run the converge procedure: read every .agents/skills/*/SKILL.md contract section, audit the repo against each bullet, fix deviations using the guidance in each skill's body, verify inside the devcontainer, and open a PR titled "Converge on dev-env standards".
AGENTS.md holds only command routing (toolchain commands run in the devcontainer, git and gh on the host), the project's always-needed context, and a skills index table. Anything needed only sometimes belongs in a skill, not in AGENTS.md.
No. Files beginning with the line "<!-- Generated by dev-env -->" are rewritten byte-for-byte on the next dev-env apply, so hand edits are silently reverted. If the generated shape is wrong, report it upstream as a dev-env bug.
No. CLAUDE.md must begin with the line @AGENTS.md and contain nothing else duplicated below it. Claude Code loads AGENTS.md through that import, so a second copy of the content just goes stale.
Cursor reads AGENTS.md natively, so a .mdc file is a second instruction source that inevitably drifts from it. The converge contract requires that no .cursor/rules/*.mdc files exist; delete any found rather than syncing them by hand.
selhorysAlign SKILL.md frontmatter and routing with current implementation state.
fagemxMonitor Pull Request CI pipelines and auto-fix linting and formatting issues.
BicameralAIBlock agent actions when bicameral detects drifted or divergent decisions.
TheBushidoCollectiveWrite portable Bash scripts with strict error handling and safe variable usage.