android-native-init-lab CLAUDE.md

Repository instructions for Temmie-Tem/android-native-init-lab that define which operating, device-safety, and target-specific rules apply. They require reading the root contract, one selected target contract, and the matching goal in order.

In plain words
What is it for?
Use them when operating in that repository to select the correct target, follow current authority limits, preserve isolation and recovery requirements, and respect attended-only device procedures.
Why use it?
They prevent work from relying on retired permissions or applying the wrong device rules, especially when working with the S22+ or A90 targets.

Instructions file

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/temmie-tem/android-native-init-lab/claude-md
Clone the repo
git clone --depth 1 https://github.com/Temmie-Tem/android-native-init-lab
Per session 497 This file is loaded in full into every session.
When invoked 497 The same file — it is already loaded in full.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00497 $0.00497
Opus 5 $0.00249 $0.00249
Sonnet 5 $0.00099 $0.00099
Haiku 4.5 $0.00050 $0.00050

Measured 2d ago against content hash 07577b086832, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

android-native-init-lab CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 40 lines

What it actually says

CLAUDE.md

AGENTS.md is the root binding operating and device-safety contract in this repository. Select exactly one binding target contract from its registry, then read AGENTS.md, that target contract, and the target goal in that order. GOAL.md defines the S22+ frontier and GOAL_A90.md defines the A90 frontier; neither goal grants device authority.

Current posture:

  • The Interim Fast-Loop Rules block at the top of AGENTS.md is RETIRED. Its retirement condition was met at 2026-08-03T20:46:02Z by the first close rows for two distinct campaign IDs. The retained trial text is historical and grants no current standing D0, procedural autonomy, or override.
  • Apply the ordinary common contract, the selected binding target contract, and its current process. Any D0, D1, or F1 authority must come from those current layers and live inputs; do not infer it from the retired trial.
  • Permanent integrity, no-replay, target isolation, recovery, and inter-effect health requirements remain mandatory.
  • Legacy v1 checks are implementation limits. The current A90 v1 runner is attended-only; never assert --operator-attended while the operator is absent.
  • A missing or late endpoint, timeout, or malformed observation freezes new device effects and enters health classification; it does not by itself close the campaign. Passive reads, host-only observer repair, and the exact predeclared recovery may continue without replaying the uncertain action.
  • Record evidence according to the ordinary AGENTS.md evidence rules and the selected target contract. The trial ledgers remain append-only historical evidence but no longer grant authority.
  • Files under docs/archive/ are historical and grant no authority, even when they contain ACTIVE, acknowledgement, or exception text.

Classify device work with docs/operations/DEVICE_ACTION_RISK_TIERS.md. Boot-only F1 design and recovery semantics are defined in docs/operations/DEVICE_ACTION_PROCESS_V2.md.

Do not reconstruct a candidate-specific policy from archived clauses. Do not contact a device when the selected unit is host-only. Permanent forbidden partitions and primitives in AGENTS.md are absolute.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 40 lines · 497 tokens per session scan A 07577b086832

Subscribe to this mod's changes

android-native-init-lab CLAUDE.md is an instructions file published in the GitHub repository Temmie-Tem/android-native-init-lab (10 stars, last pushed 4d ago), licensed MIT. It adds 497 tokens to every session, about $0.0025 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.