skills.rest
Aki2022
Aki2022COMMUNITY
Explore all skills by this author →
GitHub

origin-goal

Orchestrates long-running workstreams with human approval gates and resource ceilings.

Updated Dec 9, 2025
Search Tags:#git workflow#subagents#orchestration#autonomous agent#workstream#human approval#budget control
One-click install
npx skills add https://github.com/Aki2022/skills --skill origin-goal-aki2022
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: origin-goal
Source: https://github.com/Aki2022/skills/tree/main/origin-goal
Command: npx skills add https://github.com/Aki2022/skills --skill origin-goal-aki2022

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long-running autonomous work tends to drift: agents lose track of scope, exceed budgets, resurrect stale branches, or act without human authorization. This Skill runs a durable, workstream-driven objective inside explicit human boundaries so multi-slice work stays authorized, metered, and resumable. ## Core Features & Use Cases - Workstream Discovery and Control Units: Detects the repository's own convention (docs/workstreams, specs directories, or issue-based units) instead of imposing a parallel structure, and selects or creates the right control unit with human confirmation. - Preflight Authorization Envelope: Interviews the user one question at a time to record allowed actions, confirmation gates, resource ceilings (money, tokens, time, retries), test gates, and stop conditions before execution begins. - Subagent Orchestration and Failure Control: Routes bounded work to subagents with model-tier selection, enforces a three-strike same-root-cause stop rule, and handles stale or merged branches with deterministic salvage procedures. - Use Case: Ask the agent to keep working on a multi-issue documentation workstream until the next human checkpoint, with a token budget and PR-merge autonomy, and it will resume the recorded workstream, execute vertical slices, and stop at the defined gate. ## Quick Start Use $origin-goal to resume my active workstream and run it to the next human gate within a 200k token budget.

Frequently Asked Questions about origin-goal

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run an autonomous agent on a long task without losing control?▼

Define a workstream as the durable control unit, then complete a preflight interview recording allowed actions, confirmation gates, and resource ceilings. The agent executes vertical slices autonomously inside that envelope and stops at human gates or budget limits.

How do I resume an interrupted AI workstream in a repository?▼

The agent reads the repository's own workstream convention, verifies the recorded branch still exists, and resumes from the recorded state. Merged or stale branches are treated as bookkeeping to correct, not work to rebuild.

What happens when a recorded branch is stale or already merged?▼

A merged-and-deleted branch means continuing from the default branch. A stale unmerged branch is salvaged by reading only its own commits, then a new branch is cut from the default branch and the record is rewritten in the same slice.

How are retries and failures limited during autonomous runs?▼

Execution stops after three consecutive failures with the same root cause, counting the initial failure as attempt one. The counter resets only when evidence shows the root cause actually changed, not when the prompt wording changed.

Can the agent create and merge pull requests autonomously?▼

Yes, PR creation and merge that pass recorded test and CI gates are treated as autonomous integration by default. A human gate is recorded only when the user or repository policy explicitly requires one, and branch protection still applies.

When should I not use a workstream for a task?▼

Use a standalone issue for one bounded, reversible, non-metered change, and docs-only when nothing is implemented. Workstreams fit work spanning multiple slices, metered services, or tasks needing a human gate.

Related Skills

Explore More Software Engineering Skills→
ToXMonToXMon

a0-contribute-plugin

Publish and validate Agent Zero plugins for the Plugin Index via GitHub PR.

Community
Advanced
volter-aivolter-ai

reviewer

Reviews GitHub pull requests and publishes an agent-review verdict against policy and architecture invariants.

Official
Advanced
DagimAlemayehuuDagimAlemayehuu

triage

Triage GitHub issues through a label-driven state machine with agent briefs.

Community
Intermediate
davepoondavepoon
3.4k

checkpointed-agent-loop

Runs resumable Claude Code tasks as bounded loops with checkpointed state and verification evidence.

Community
Advanced