orbit_skill

This skill stabilizes autonomous loops by enforcing contracts, validating artifacts, and guiding safe, reversible next actions for Nexus workflows.
  • Shell

8

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill simota/agent-skills --skill orbit

  • SKILL.md27.6 KB

Overview

This skill enforces loop-operation contracts for nexus-autoloop style autonomous workflows. It focuses on making autonomous loops predictable, auditable, and safely recoverable by validating loop artifacts and governing status semantics. Orbit produces evidence-first recommendations and minimal reversible actions to stabilize or advance loops.

How this skill works

Orbit parses required artifacts (goal.md, progress.md, state.env) and optional evidence (done.md, runner.log), validates the NEXUS_LOOP_STATUS/NEXUS_LOOP_SUMMARY footer, and scores evidence quality. It classifies failure modes (verify gaps, state drift, commit-scope risk), generates a severity-backed next action, and prepares rollback-safe handoffs to Builder, Guardian, Radar, or Nexus. High-impact decisions trigger AskUserQuestion prompts before destructive changes.

When to use it

  • When a loop claims DONE but verification evidence is missing or failed
  • When resuming a loop and state.env counters disagree with progress.md
  • Before recommending an autocommit when the working tree may be dirty
  • When goal acceptance criteria are vague or not measurable
  • When you need a minimal reversible fix to unblock a stalled loop

Best practices

  • Require measurable acceptance criteria in goal.md before declaring DONE
  • Prefer the smallest reversible change and document rollback notes
  • Separate contract violations (status/footers) from implementation bugs
  • Always include verification commands and outcomes with any DONE claim
  • Ask for explicit user confirmation before discarding or rewriting user changes

Example use cases

  • Audit a loop that reported DONE but lacks verification output, returning CONTINUE with a remediation checklist
  • Detect state drift on resume and produce a re-sync plan based on progress.md iteration evidence
  • Limit autocommit scope when dirty baseline files are present and hand off safe staging to Guardian
  • Create an Orbit→Builder handoff that specifies exact files, failure class, and rollback steps for a minimal patch
  • Operate as the completion gate to downgrade false DONE claims and route verification work to Radar

FAQ

By default Orbit recommends CONTINUE and provides a prioritized evidence checklist and the smallest reversible next steps to reach verifiable DONE.

Will Orbit ever rewrite progress.md or done.md?

No. Orbit does not rewrite operational history without explicit user confirmation; it may propose edits or annotate inconsistencies but will ask before making changes.

How does Orbit handle autocommit risk when the baseline is dirty?

Orbit classifies COMMIT_SCOPE_RISK, restricts staged paths, recommends path-scoped commits, and triggers an ON_AUTOCOMMIT_SCOPE_RISK confirmation when ambiguous.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
orbit skill by simota/agent-skills | VeilStrat