[PR #941] [CLOSED] Add ASIL: autonomous improvement loop with cost control and adversarial gates #14089

New Issue

GiteaMirror · 2026-07-11T19:31:59-05:00

GiteaMirror commented

2026-07-11 19:31:59 -05:00

📋 Pull Request Information

Original PR: https://github.com/Shubhamsaboo/awesome-llm-apps/pull/941
Author: @telivity-otaip
Created: 7/10/2026
Status: ❌ Closed

Base: main ← Head: add-asil-external-listing

📝 Commits (1)

c2d543a Add ASIL as external advanced agent and cost-control listing

📊 Changes

1 file changed (+2 additions, -0 deletions)

View changed files

📝 README.md (+2 -0)

📄 Description

Summary

Adds ASIL (Autonomous Software Improvement Loop) as an external listing in two sections where it fits the bar:

Advanced AI Agents — production autonomous coding pipeline (scan → fix → three-persona self-review → adversarial gate → domain guard → open PR)
LLM Optimization Tools — first-class cost controller (per-task budgets, call-level checkpoints, daily kill switch, spend attribution)

ASIL is the open-source extract of a system that has been running unattended in production (80+ agent travel AI platform), opening PRs autonomously with human-meaningful improvements landing in main.

Why this complements the existing catalog

Shubham — built a self-improvement loop using agent skills patterns that works well across repos, with a more adversarial take on cost control and conflict resolution. ASIL ships the pieces your advisor-orchestrator-worker skill points at but as runnable infrastructure:

Pattern in awesome-llm-apps	ASIL implementation
Advisor consulted at commitment boundaries	Papa agent resolves synthesizer conflicts and produces handoff briefs
Orchestrator dispatches stateless workers	Thought Multiplier router → specialized thinkers → synthesizer
Budgeted worker calls	Cost Controller (System C) — every LLM call passes through a checkpoint; refuses when budget is exhausted
Verification before merge	Three-persona self-review + separate-LLM adversarial gate before any PR opens
—	Domain guard — blocks autonomous fixes on `DOMAIN_QUESTION:` zones until a human triages

Designed for open-model orchestration + SLM specialist workers (Ollama, vLLM, LM Studio via OpenAI-compatible adapter) with edge/AWS deployment paths — avoids bleeding tokens on frontier models where domain knowledge should stay proprietary.

Listing format

Follows the existing external pattern (Openwork, Jarvis) — links out to the standalone repo rather than vendoring a TypeScript monorepo into this cookbook.

Test plan

README renders correctly (two new bullet entries, external badge)
Links resolve to https://github.com/TelivityAI/asil
Maintainer review for placement / Featured consideration

Happy to follow up with a native agent_skills/ contribution (Papa + cost-control worker skill) if you'd prefer that over an external listing.

— Dušan Milicevic / Telivity

Made with Cursor

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/Shubhamsaboo/awesome-llm-apps/pull/941 **Author:** [@telivity-otaip](https://github.com/telivity-otaip) **Created:** 7/10/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `add-asil-external-listing` --- ### 📝 Commits (1) - [`c2d543a`](https://github.com/Shubhamsaboo/awesome-llm-apps/commit/c2d543a964bc5cd28f8f8192c50fa514470f9b87) Add ASIL as external advanced agent and cost-control listing ### 📊 Changes **1 file changed** (+2 additions, -0 deletions) <details> <summary>View changed files</summary> 📝 `README.md` (+2 -0) </details> ### 📄 Description ## Summary Adds [ASIL (Autonomous Software Improvement Loop)](https://github.com/TelivityAI/asil) as an external listing in two sections where it fits the bar: - **Advanced AI Agents** — production autonomous coding pipeline (scan → fix → three-persona self-review → adversarial gate → domain guard → open PR) - **LLM Optimization Tools** — first-class cost controller (per-task budgets, call-level checkpoints, daily kill switch, spend attribution) ASIL is the open-source extract of a system that has been running unattended in production (80+ agent travel AI platform), opening PRs autonomously with human-meaningful improvements landing in main. ## Why this complements the existing catalog Shubham — built a self-improvement loop using agent skills patterns that works well across repos, with a more adversarial take on cost control and conflict resolution. ASIL ships the pieces your advisor-orchestrator-worker skill points at but as runnable infrastructure: | Pattern in awesome-llm-apps | ASIL implementation | |---|---| | Advisor consulted at commitment boundaries | **Papa agent** resolves synthesizer conflicts and produces handoff briefs | | Orchestrator dispatches stateless workers | **Thought Multiplier** router → specialized thinkers → synthesizer | | Budgeted worker calls | **Cost Controller** (System C) — every LLM call passes through a checkpoint; refuses when budget is exhausted | | Verification before merge | **Three-persona self-review** + separate-LLM **adversarial gate** before any PR opens | | — | **Domain guard** — blocks autonomous fixes on `DOMAIN_QUESTION:` zones until a human triages | Designed for open-model orchestration + SLM specialist workers (Ollama, vLLM, LM Studio via OpenAI-compatible adapter) with edge/AWS deployment paths — avoids bleeding tokens on frontier models where domain knowledge should stay proprietary. ## Listing format Follows the existing external pattern (Openwork, Jarvis) — links out to the standalone repo rather than vendoring a TypeScript monorepo into this cookbook. ## Test plan - [x] README renders correctly (two new bullet entries, external badge) - [x] Links resolve to https://github.com/TelivityAI/asil - [ ] Maintainer review for placement / Featured consideration Happy to follow up with a native `agent_skills/` contribution (Papa + cost-control worker skill) if you'd prefer that over an external listing. — Dušan Milicevic / [Telivity](https://telivity.com) Made with [Cursor](https://cursor.com) --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

GiteaMirror added the pull-request label 2026-07-11 19:31:59 -05:00

GiteaMirror closed this issue

2026-07-11 19:32:00 -05:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/awesome-llm-apps#14089