llm-grace/0-AI-MANIFEST.a2ml at main · hyperpolymath/llm-grace · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
# SPDX-License-Identifier: MPL-2.0
# Copyright (c) 2026 Jonathan Jewell (hyperpolymath) <j.d.a.jewell@open.ac.uk>
#
# AI-MANIFEST.a2ml — AI agent work allocation and policy
#
[metadata]
version = "0.1.0"
last-updated = "2026-05-18"

[project]
name = "llm-grace"
purpose = "Graceful degradation for concurrent LLM/agent terminals: cheap signature-based load-shedding with a crash-safe ledger and legible post-mortems"

[ai-allocation]
agents = [
  { name = "CLAUDE", role = "proofs, compilers, repo-local implementation, CI/CD, formal verification, Rust/Idris2/Zig" },
  { name = "CHATGPT", role = "prose, papers, publication review, standards docs, outreach drafts, letters" },
  { name = "GEMINI", role = "estate audits, cross-repo sweeps, long-context triage, pattern detection" },
  { name = "VIBE", role = "UI/frontend, PanLL panels, ReScript components, theming, rapid prototyping" },
]

[policy]
rules = [
  "Do not duplicate tasks across sections. If a task needs multiple LLMs, note the handoff.",
  "Update THIS file during sessions. Do NOT recreate per-repo TODO files.",
]

[work-allocation]
items = [
  { agent = "CLAUDE", task = "proofs, compilers, repo-local implementation, CI/CD, formal verification, Rust/Idris2/Zig" },
  { agent = "CHATGPT", task = "prose, papers, publication review, standards docs, outreach drafts, letters" },
  { agent = "GEMINI", task = "estate audits, cross-repo sweeps, long-context triage, pattern detection" },
  { agent = "VIBE", task = "UI/frontend, PanLL panels, ReScript components, theming, rapid prototyping" },
]