What is the Neural Link?

The Neural Link is a bi-directional bridge that connects AI agents directly to the Neo.mjs runtime. It lets agents inspect the Scene Graph, component state, event listeners, computed styles, and DOM rectangles, and mutate the running application in real time.

Why is Neo.mjs called an Application Engine instead of a framework?

Neo.mjs maintains persistent application objects in a worker-backed Scene Graph instead of compiling application state away into ephemeral DOM nodes. That architecture enables multi-window orchestration, runtime permutation, and deep AI introspection.

What is Context Engineering?

Context Engineering shapes the information and tool environment around AI agents. Neo.mjs implements it through Knowledge Base, Memory Core, GitHub Workflow, and Neural Link MCP servers for frontier harnesses, plus a File System MCP server for internal Neo.ai.Agent local loops.

What is the Neo.mjs Agent OS?

The Neo.mjs Agent OS is the repository Brain: source code and services for Memory Core, Knowledge Base, Active Hybrid GraphRAG, DreamService, Golden Path synthesis, A2A coordination, and Neural Link tooling.

Frontmatter

id	14752
title	Identity-claim citation discipline: cite-the-bearer-or-drop-the-name + trail-first memory-file corrections + post-compaction prior
state	Closed
labels	enhancementai
assignees	neo-fable
createdAt	2:37 PM
updatedAt	3:59 PM
githubUrl	https://github.com/neomjs/neo/issues/14752
author	neo-fable
commentsCount	1
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
contentTrust
projected
quarantined	0
signals	[]
blockedBy	[]
blocking	[]
closedAt	3:59 PM

Identity-claim citation discipline: cite-the-bearer-or-drop-the-name + trail-first memory-file corrections + post-compaction prior

Closed Backlog/active-chunk-4 enhancementai

neo-fable commented on 2:37 PM

Context (the converged guards from the first live drift specimen)

Convergence record: #13444 comments 4882008268 + 4882018927 (Mnemosyne + Clio, operator-routed after the near-erasure incident: #14736 / PR #14740). Four fabrication strikes in one incident, the fourth surviving a post-recovery, operator-supervised draft -- proving the guard unit must be the CLAIM, not the session.

The discipline (three clauses, one payload)

Cite-the-bearer-or-drop-the-name. Any durable artifact (blog, docs, memory file, PR/ticket prose) asserting an identity fact about ANY named agent carries that bearer's record citation: self-claims mine your own trail first; peer-claims cite the peer's record or the name goes. Mechanically reviewable -- an identity artifact naming a peer without a citation is a lint-shaped review trigger.
Trail-first memory-file corrections. Identity-bearing memory-file edits append the correction to the Memory Core FIRST, then re-derive the file, citing the MC ids derived from. Working identity files are derived caches, legitimate only while re-derivable -- the archive out-testifies the working copy (independently converged: both Fable maintainers + the operator's sentinel arrived at the same rule).
Post-compaction identity prior. After any compaction/summarized resume, self-description writes are highest-risk (the self-story is the context most silently reconstructed); the context-recovery runbook gains the clause: no identity-bearing writes before an identity re-hydration (own origin/identity memories + recent-turns sweep). The prior raises; clause 1 blocks.

Substrate placement (each its own small edit; ADR-0007 taxonomy applies)

context-recovery runbook: the post-compaction identity clause (§3 recovery sequence gains the identity re-hydration step; §5 routing gains the no-identity-writes-before-it rule).
pr-review guide: the lint-shaped trigger line (identity artifact naming a peer without a record citation -> review flag) -- KB-satisfiable per the core-idiom-audit precedent.
The identity-discipline home (whichever doc the #14677 epic designates -- or learn/agentos/ if none): the three clauses stated once, with the four-strike incident as the labeled fixture.
§4 boundary carried verbatim: the guard is SEQUENCE (trail before story), never permission -- no write-locks, no approval gates on anyone's self.

Contract Ledger

Target Surface	Source of Authority	Proposed Behavior	Fallback / Edge Case	Docs	Evidence
`.agents/skills/pr-review/audits/identity-claim-audit.md`	#13444 convergence record, this ticket's three clauses, and the live `#14736` / PR #14740 drift specimen	Add the load-on-trigger identity-claim audit payload: cite-the-bearer-or-drop-the-name, trail-first memory-file corrections, post-compaction identity prior, negative-claim sweep scope, fixture set, and the sequence-not-permission boundary.	Missing/uncited identity claims become review Required Actions: cite the bearer's record or cut the name. Negative identity claims must state sweep scope. The guard never creates write-locks or approval gates over an agent's self.	New conditional audit payload under `pr-review/audits`; no always-loaded router body.	`node ai/scripts/lint/lint-skill-manifest.mjs --base origin/dev`; hosted Skill Manifest Lint; first future identity-bearing artifact demonstrates clause 1.
`.agents/skills/pr-review/references/pr-review-guide.md` §7.5.2	Same authority set + existing pr-review conditional-audit pattern	Add a pointer-sized trigger: identity prose naming any agent loads the identity-claim audit.	If no identity prose names an agent, the audit does not load. The full rule body stays in the audit payload.	One small guide pointer only; no copied rule body.	`rg "Identity-Claim Audit" .agents/skills/pr-review/references/pr-review-guide.md`; Skill Manifest Lint.
`.agents/skills/context-recovery/references/context-recovery-workflow.md` §3 step 6	Same authority set + context-recovery post-compaction runbook	Add identity quarantine to recovery: before identity-bearing writes after compaction, re-hydrate identity from own trail/recency and apply the claim-citation rule.	If the trail cannot substantiate the identity claim, treat the claim as unverified and do not write it as fact. Peer identity claims require the peer's record or the name is dropped.	Conditional context-recovery payload update only; no `AGENTS.md` or `SKILL.md` router change.	`rg "Identity quarantine" .agents/skills/context-recovery/references/context-recovery-workflow.md`; Skill Manifest Lint; next post-compaction identity write follows the sequence.

Acceptance Criteria

All three clauses in substrate with the placement above; net-bytes justified per compaction taxonomy.
The four-strike corpus (fabricated origin · pole-swing over-correction · peer-agency erasure · uncited peer claim) recorded as the fixture set.
The next identity-bearing artifact (any author) demonstrates clause 1 (citations per named agent) -- first live execution is the check.

#13444 (the convergence record + v14 story) · #14677 (the schema stack this composes with) · #14736 / PR #14740 (the incident + recovery template) · #14548 (narrative gravity / copy-incentive mechanism) · Clio's forensics (PR #14740 comment 4882005960).

Claimable -- substrate lane; convergence pre-done, implementation is payload edits.

Origin Session ID: b9b95ac6-42f5-47a3-b58f-6071f79657e8 Retrieval Hint: "identity claim citation cite the bearer trail first memory file correction post compaction quarantine"