next-recursive-prompt.uai

Carries the sectioned next-loop resume prompt for code-bearing agents without replacing authoritative .uai records or current human instruction.

What this file is#

A derived hot-memory resume artifact using canonical JSON profile uaix.next-recursive-prompt.v1. It is not a second source of truth.

Why necessary#

Recursive coding sessions often lose the exact next work chunk after compaction or handoff. This file keeps the next loop executable while forcing authoritative records to win on conflict.

Activation rule#

• Use `.uai/next-recursive-prompt.uai` when the current request is continuation, resume, keep-going, do-the-next-thing, proceed, or directly related to the stored next loop.
• Review related plan ideas before discarding or superseding the stored next loop.
• If the human gives a specific unrelated instruction, the current instruction governs and the file is regenerated after accepted work.
• Related active File Handoff intake supersedes or expands the stored loop and still requires project work, ledger evidence, configured durable-memory preservation, and source-site removal before completion; retained files are unfinished human-hold or blocker states.
• At the end of accepted work, show the human a prominent, sectioned, detailed suggested next prompt that they can review, modify, replace, or answer with keep going.

Dogfood writeback rule#

• UAIX.org is code-bearing, so its local `.uai/next-recursive-prompt.uai` must be present, current, and treated as derived hot-memory only.
• After accepted UAIX source work, update authoritative `.uai` records first, then regenerate or refresh the next-recursive prompt from those records.
• The file must point to current checks and blockers without copying secrets, raw logs, change-controlled anchor contents, or stale chat history.
• The visible next prompt must be a standalone, copy-paste-ready handoff and must use literal Feature/improvement focus, Code/source changes, and Automated tests/checks sections with actionable detail, plus context, objective, scope, hard boundaries, work items, expected touchpoints, files to review, done criteria, and final response requirements.

When required#

Required for agents that deal with code, automation, executable workflows, or code-like artifacts: next-recursive-prompt.uai..

What belongs#

• The next safe work chunk and acceptance evidence.
• Authoritative source references to read before acting.
• Checks to run and blocked-recovery behavior.
• Refresh triggers and writeback targets for the next loop.

What does not belong#

• Secrets, credentials, access tokens, private keys, or private incident details.
• Durable project truth that belongs in typed active `.uai` records such as progress, decisions, constraints, test-plan, stack, or architecture records.
• Raw chat logs, broad speculation, or stale history that belongs in archives.
• A YAML mirror or alternate source of truth.

How receiving AI should use it#

• Read authoritative_refs before using the prompt.
• Use the stored plan when the user asks to continue, keep going, do the next thing, resume, proceed, or otherwise continues the recorded next loop.
• When the current request directly relates to the stored next loop, review this file for useful plan ideas before discarding or superseding it.
• When the user gives a specific unrelated instruction, current user instruction governs the turn and this file is regenerated after accepted work.
• Treat the file as derived hot-memory only.
• If it conflicts with authoritative .uai records, stop and regenerate it from those records.
• Fail closed or enter blocked recovery when the file is stale, invalid, contradictory, or points to missing evidence in a required code-bearing scope.
• End accepted continuation work with a prominent, sectioned, detailed suggested next prompt in chat so the human can review, modify, replace, or answer with keep going.

How human reviewer should use it#

• Confirm the file is JSON with profile uaix.next-recursive-prompt.v1.
• Confirm it names the exact next work chunk, checks, blockers, and update targets.
• Confirm the final report shows a prominent, copy-paste-ready Suggested Next Prompt instead of only storing the next loop in memory. The prompt must include literal Feature/improvement focus, Code/source changes, and Automated tests/checks sections with actionable detail.
• Confirm no secret or private credential is present.
• Confirm it does not override authoritative .uai records.

Scenario examples#

• After a long coding pass, regenerate the file so the next agent can continue without replaying the full transcript.
• When a human says keep going, use it to recover the next work chunk after reviewing authoritative memory.
• When a human wants review control, show the next detailed suggested prompt in a visible sectioned block. It may shorten low-value repetition only when it remains copy-paste-ready and preserves literal Feature/improvement focus, Code/source changes, and Automated tests/checks sections with actionable detail before continuation.
• When a human asks for an unrelated task, review it only for relevant context, then follow the current request and regenerate afterward.
• After a failed test run, record the blocker and the next recovery step.
• After processing active intake, point the next loop at the updated ledger and remaining work.

No-op and human-review triggers#

• The file is required but absent from a code-bearing package.
• The file is not valid JSON or does not use profile uaix.next-recursive-prompt.v1.
• The next-loop prompt contradicts the active `.uai/` startup suite, progress, constraints, coding standards, stack, architecture, or test-plan.
• The file contains secrets or unsupported public claims.

Update and maintenance rules#

• Regenerate after meaningful work, test results, blockers, deployment, handoff, intake processing, and context compaction.
• Keep bulky history in archives and preserve only the current next-loop slice here.
• Write accepted current truth into authoritative .uai records first, then regenerate this file from those records.

Relationship to other .uai files#

• Derived from AGENTS.md, memory-maintenance, identity, world-context, Totem, Taboo, Talisman, startup packet, system profile, receiver brief, coding standards, stack, architecture, context, constraints, progress, next actions, test plan, and manifest.
• The active `.uai/` startup suite keeps current truth; this file keeps the next recursive prompt.
• Progress and test-plan carry evidence; this file points to the next action and checks.

Minimal example#

# next-recursive-prompt.uai
file: .uai/next-recursive-prompt.uai
required_status: Required for agents that deal with code, automation, executable workflows, or code-like artifacts: next-recursive-prompt.uai.
owner: human-reviewed owner or steward
current_state: accepted current record for this package scope
no_op_trigger: Missing, stale, invalid, contradictory, or evidence-broken next-recursive prompt in a required code-bearing scope.
evidence: link or local pointer to reviewed source
review_status: current until superseded by owner review

Completeness checklist#

• Top-level fields match uaix.next-recursive-prompt.v1.
• Authoritative refs are present and current.
• Next loop names a concrete work chunk, checks, blockers, and writeback policy.
• No secrets, private credentials, unsupported claims, or stale instructions are present.

Common mistakes#

• Using it as a second source of truth.
• Letting it override a newer unrelated human instruction.
• Discarding a related stored plan without reviewing it for useful work ideas.
• Letting it survive after authoritative records changed.
• Storing a full chat transcript instead of a compact next-loop prompt.
• Adding it to no-code assistant packages before code, automation, executable workflow, or code-like artifacts are declared or detected.

Related UAIX records#

• AI Memory
• AI Memory Package Wizard
• Project Handoff
• Agent File Handoff
• AGENTS.md Specification

Machine-readable digest#

Agents should treat this digest as page-orientation evidence, not runtime authority.