AGENTS.md git:20260930.19be338A

Immutable. This exact content never changes and is served at /api/v1/blob/6dd0460f4063d331.

# AI work instructions

Follow [CONTRIBUTING.md](CONTRIBUTING.md) for required reading, verification,
release preparation, and submission.

Carry an action request through to its intended result. Use the conversation's
existing decisions and authorization; choose routine, reversible details
without asking again. Ask only when missing information changes the result or
an action needs authority the person has not given. Continue independent,
authorized work while an answer is pending.

Explicit user instructions take precedence over Skill guidance. If a Skill
would pause or redirect authorized work, first check whether that rule applies.
When it does, quote and link the exact instruction and explain the remaining
decision. Prepare the authorized work so any needed approval concerns a
concrete, reviewable result.

Apply the shared [writing standard](plugins/hope/skills/write/references/writing-standard.md).
Invoke Hope skills when the person explicitly selects them, with two exceptions:
PR Writing applies when drafting or updating PR/MR titles and descriptions,
including during an authorized submission workflow; it calls Write through
their public handoff. Keep updates and results concise; report the change,
relevant evidence, and any verification gap.

These collaboration rules draw on the
[GPT-6 Astra prompting guide](https://developers.openai.com/api/docs/guides/latest-model?model=gpt-6-astra#prompting-best-practices).