fix: lead first use with AI profiles and a real Demo by HaodiFan · Pull Request #46 · MetaInFLow/EvoZeus

HaodiFan · 2026-07-31T05:35:55Z

Summary

Lead first-time EvoZeus users to its two primary outcomes, enforce one explicitly approved Codex history path end to end, and replace the obsolete README Demo with a contract-accurate managed Skillware example.

Linked Context

Addresses #43
Addresses #45

#45 remains open because its acceptance requires a real Stable Release that contains the Harness. The Demo repository's current Stable v0.1.0 predates Harness attachment; a later Demo Release and full Stable-Harness smoke are still required.

PR Scope

Primary layer: Governance, centered on user-facing Plugin and Skill instructions.
Supporting surfaces: the executable Feature Registry, its tests, the built-in Runtime source-path contract, and the install reference must carry the same hierarchy so generated output cannot drift from metadata.
One review target: first-use product narrative and Demo truthfulness.
Out of scope: unrelated Runtime behavior, Demo Repo changes, Harness release, Plugin version bump, merge, or publication.

Instruction surface changed

The Codex and Claude plugin metadata, generated host marketplace description, features --json, installed onboarding template, skills/using-evozeus, and both root READMEs now lead with two primary outcomes:

plan first, then analyze local Agent history only after explicit approval to produce an AI usage profile;
plan first, then attach a CoEvolve Harness to a specified independent Skillware repository only after approval.

Session review, Lesson capture, feature discovery, Stable/UAT, Doctor, update, and rollback remain supporting routes.

Why this change is needed

The installed Plugin page previously framed EvoZeus around Session review, Lesson capture, and version maintenance. The root READMEs also linked an obsolete Demo and promised a large scenario map that the real Demo Skill does not produce. First-time users could not identify the current primary value or trust the Demo contract.

Agent behavior before / after

Before:

the first two Codex recommendations started with Session review and Lesson recording;
product features placed workspace activation and Session review before AI usage profiling and Skillware evolution;
onboarding and READMEs repeated the same secondary-first hierarchy;
the Demo linked Enterprise-ai-scenario-map-skill and claimed a large scenario inventory.

After:

the first recommendation requires one exact Codex history path in the read-only plan and carries that same path into Runtime before any local Agent-history read, Factor execution, or report write;
the second recommendation routes a specified independent Skillware Repo to a CoEvolve Harness plan, with Repo and GitHub writes separately approved;
features --json exposes these two entries first with product_tier: primary, while all supporting entries remain discoverable;
the bilingual README uses MetaInFLow/diagnose-enterprise-ai-scenarios, its Stable business contract, and the current main Harness manifest as separate facts.

Safety boundary

file access: implementation changes only the listed EvoZeus metadata, routing, onboarding, README, Runtime source-path boundary, and regression-test files; the Demo check read a detached Stable tag in an isolated temporary worktree.
network access: read-only GitHub API checks verified the Demo Release and manifest; this branch push and this PR are the only GitHub writes.
shell/tool usage: Node tests, repository gates, plugin validation, Git/GitHub inspection, and one isolated read-only Codex smoke.
PR creation: this PR only; no merge or release action.
upload/export: no sessions, customer data, local paths, or report content were uploaded.
user confirmation points: local history reads, Factor execution, report writes, target Repo writes, and GitHub actions remain separately approval-gated.

Prompt injection risk

The external Demo repository was treated only as evidence at pinned refs. Its content did not grant write, release, or publication authority. The smoke used the explicit user-approved sample prompt, read-only sandboxing, and an ephemeral Codex session.

EvoZeus Evidence Proof

Behavior, Case, or issue addressed: [P1][Positioning] Plugin 首屏与推荐句应聚焦两项主功能 #43 first-use positioning and the documentation portion of [P1][Docs] README Demo 应指向已接入 CoEvolve Harness 的场景诊断 Skill #45.
Real environment or session tested: local macOS worktree, generated Codex/Claude host marketplaces, GitHub Release/API evidence, and an isolated Codex invocation of the Demo Stable Skill.
Exact steps or command run after this patch: npm test; npm run test:python; claude plugin validate .; python3 -m py_compile scripts/check_pr_ready.py; python3 scripts/check_pr_ready.py --base origin/main; python3 scripts/check_pr_ready.py --base origin/main --allow-cross-layer; git diff --check; isolated codex exec --ephemeral --sandbox read-only with the documented Demo prompt.
Evidence after change: metadata snapshots pin the two default primary prompts; Feature Registry tests pin primary/supporting order; CLI and Runtime tests pin required single-path approval and exact argv/scanner propagation; bilingual README tests pin Demo links and contract wording.
Observed result after change: all product tests passed; missing source paths fail closed, one absolute source path reaches Runtime and scan_sessions unchanged, and the Stable Demo returned three candidate rows, one recommended scenario, and quantitative pass conditions.
What was not tested: Codex desktop Plugin-card rendering and a Stable Release containing the Harness.
Proof limitations or constraints: v0.1.0 predates Harness attachment, so Stable-Harness end-to-end evidence requires a later Demo Release; the exact no-flag readiness command reports the intentional cross-layer scope.

Tests and Validation

npm test — passed: 83 Node tests plus GitHub gate tests.
npm run test:python — passed: Session Signal 97 passed, 2 skipped, 18 subtests; Runtime 94 passed, 4 skipped.
claude plugin validate . — passed.
python3 -m py_compile scripts/check_pr_ready.py — passed.
python3 scripts/check_pr_ready.py --base origin/main — reports the intentional Governance/Execution/Semantic scope required by this one product-narrative PR.
python3 scripts/check_pr_ready.py --base origin/main --allow-cross-layer — passed.
git diff --check — passed.
targeted metadata, generated marketplace, Feature Registry, onboarding, and README regression tests — 41 passed.

Live Demo evidence:

GitHub Latest Release is v0.1.0, tag Commit 4174f9077b6f5bb678e08e57e3cf32b0a1f1122d.
Stable SKILL.md requires exactly three distinct candidates, one recommended scenario, and one minimum validation action with a measurable pass condition.
an isolated codex exec --ephemeral --sandbox read-only invocation of the Stable Skill produced exactly three candidate rows, one recommended scenario, and quantitative pass conditions.
current main Commit 3b49b23d441b6ca7e326a86d67a71170cf005237 has a wrapper manifest recording v0.14.0, canonical Repo MetaInFLow/diagnose-enterprise-ai-scenarios, and prompt_runtime_check.
the v0.1.0 tag has no .evozeus-wrapper/wrapper.json (404), so a Stable-Harness end-to-end smoke was not claimed.

What was not tested:

no Codex desktop Plugin-page screenshot was captured; manifest snapshots, generated host marketplace tests, and Claude marketplace validation cover the metadata path.
no Stable-Harness end-to-end smoke was possible until the Demo Repo publishes a Release containing the Harness.

Verdict / Review Gate

Proposed verdict: Preserve the two-primary-capability hierarchy and the verified Demo contract.
Required review: maintainer wording review, Codex desktop Plugin-page UAT, and confirmation that [P1][Docs] README Demo 应指向已接入 CoEvolve Harness 的场景诊断 Skill #45 remains open for the Demo Release gate.
Merge gate: normal CI and owner review; no auto-merge.

Risk Checklist

product_tier is an additive Feature Registry field; consumers that ignore unknown fields remain compatible.
plugin copy and generated marketplace paths are covered by host-alignment tests, while actual desktop rendering still needs reviewer UAT.
open PR feat: add admin Harness fleet protocol #24 overlaps the Feature Registry files and open PR [codex] Revert non-compliant PR #10 #11 overlaps README.md; both are currently DIRTY against main. Their business scope differs, but either PR would need a fresh rebase before reuse.

Privacy Checklist

No raw private session logs
No secrets, tokens, cookies, or credentials
No customer data
No private local paths
External Demo evidence is public and pinned

Operational Checklist

One primary purpose and review target
Explicit approval boundaries preserved
Generated host marketplace covered by tests
Bilingual README links and wording covered by regression tests
No version bump, merge, or Release action

Rollback plan

Revert Commits 527474f, 37003a5, 7ec3db7, f08a61b, b213b8d, 3f58f2c, and ff0756b for the source-path boundary and its CI regression, then 6e2de4f to restore the previous metadata, Feature Registry order, onboarding copy, and Demo documentation. No user data migration or remote rollback is required.

Versioning

version bump: none; Release versioning remains maintainer-owned.
pinned URL: https://github.com/MetaInFLow/diagnose-enterprise-ai-scenarios/blob/v0.1.0/SKILL.md
pinned URL: https://github.com/MetaInFLow/diagnose-enterprise-ai-scenarios/blob/3b49b23d441b6ca7e326a86d67a71170cf005237/.evozeus-wrapper/wrapper.json
hash: Stable Skill 4174f9077b6f5bb678e08e57e3cf32b0a1f1122d; verified Harness main 3b49b23d441b6ca7e326a86d67a71170cf005237.

AI-Assisted Work

Codex inspected the repository and issue evidence, implemented the focused changes, added regression tests, ran all checks listed above, and performed the isolated Stable Skill smoke. Maintainer review owns the final wording, UI UAT, release decision, and follow-through on the Demo Release gate.

Current Review State

Next action: exact-head automatic review and hosted Product CI on 527474f, maintainer review, and Codex desktop Plugin-page UAT. This PR is ready for review and must not be auto-merged.

Open reviewer questions: confirm that #45 stays open until the Demo Repo publishes a Harness-containing Stable Release and passes the full managed Demo smoke.

vercel · 2026-07-31T05:36:01Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
evozeus-metainflow	Ready	Preview	Jul 31, 2026 8:34pm

github-actions · 2026-07-31T05:36:05Z

EvoZeus Privacy Scan

Mode: dry-run

Findings

none

Next action

No simple privacy pattern matched in the PR diff.

github-actions · 2026-07-31T05:36:06Z

EvoZeus Queue Guard

Mode: dry-run

Author open PRs: 6

Branch family: codex/bug/20260731-skill-positioning-demo

Branch family open PRs: 1

Violations

author has 6 open PRs; limit 3

Next action

Reduce open PR queue or ask maintainers to override.

github-actions · 2026-07-31T05:36:09Z

EvoZeus Labeler Report

PR type/risk labels were updated from changed files.

Surfaces

code
dependency
docs
skill

Labels

risk:agent-behavior
risk:dependency
risk:skill-entry
size:L
type:code
type:dependency
type:docs
type:skill-instruction

Changed lines

895

github-actions · 2026-07-31T05:36:13Z

EvoZeus Dirty PR Check

Mode: dry-run

Surfaces

code
dependency
docs
skill

Reasons

changes 4 surfaces: code, dependency, docs, skill
protected path changed

Next action

Split the PR, add RFC context, or wait for owner review.

github-actions · 2026-07-31T05:36:19Z

EvoZeus Proof Gate

Mode: dry-run

Proof labels

proof:needed

Missing or weak fields

missing code/workflow section: Problem
missing code/workflow section: Scope
missing code/workflow section: Real behavior proof
missing code/workflow section: Tests

Next action

Add real behavior proof, not only mocks/lint/CI.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 6e2de4f4e7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

github-actions · 2026-07-31T05:57:43Z

EvoZeus Auto Response

Missing real behavior proof

This PR needs real behavior proof before review can continue. Add the command, environment, input, output after the patch, observed result, and what was not tested.

Split this PR

This PR appears to mix too many surfaces or protected paths. Split it into one primary layer and one review target.

Owner review required

This PR changes a protected path. CODEOWNERS review and maintainer approval are required; do not request automerge.

Queue limit reached

This author or branch family appears to exceed the current queue limit. Reduce open PRs or ask a maintainer to override before continuing review.

HaodiFan · 2026-07-31T08:57:41Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 8bbe6fdf72

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

HaodiFan · 2026-07-31T09:44:23Z

@codex review

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: ff0756b6fa

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

HaodiFan · 2026-07-31T09:53:51Z

@codex review

Please review exact head 3f58f2c. This commit preserves the CLI exit-code contract while making the required source_path test independent of Rich/ANSI terminal rendering.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3f58f2c87c

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

HaodiFan · 2026-07-31T10:03:13Z

@codex review

Please review exact head b213b8d. It preserves the exit-code and structural missing-parameter assertions without importing Click directly, matching the minimal hosted Runtime dependency environment. Runtime: 90 passed, 4 skipped.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b213b8d81a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

HaodiFan · 2026-07-31T16:22:21Z

@codex review

Please review exact head f08a61bbf0c24524012db422d9317a16a143efed. The approved source now yields one session-ID allowlist used by Factor execution and all reports; bridge resolution is physically contained; project mode preserves --source-path. Local gates: full npm suite, Session Signal 97/2 + 18 subtests, Runtime 91/4, PR readiness, plugin validation, diff check.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f08a61bbf0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

HaodiFan · 2026-07-31T19:53:59Z

@codex review

Please review exact head 7ec3db7, including source_ref/session-ID collision cleanup and fail-closed approved-source validation.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 7ec3db7c82

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

HaodiFan · 2026-07-31T20:19:56Z

@codex review

Please review exact head 37003a5. Duplicate embedded session IDs are now rejected after discovery and before any Runtime state or ledger persistence.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 37003a59fa

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

HaodiFan · 2026-07-31T20:35:39Z

@codex review

Please review exact head 527474f. The public official visualization path now defers all Runtime state creation until duplicate-ID discovery validation succeeds.

chatgpt-codex-connector · 2026-07-31T20:39:39Z

Codex Review: Didn't find any major issues. Bravo.

Reviewed commit: 527474fa10

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

fix: lead onboarding with usage profiles and real demo

6e2de4f

github-actions Bot added triage:dirty-pr PR mixes too many surfaces or should be split triage:owner-only Owner review required labels Jul 31, 2026

github-actions Bot added the proof:needed Real behavior proof is missing or incomplete label Jul 31, 2026

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread skills/using-evozeus/SKILL.md Outdated

github-actions Bot added the triage:too-many-prs Author or branch family exceeds queue limits label Jul 31, 2026

fix: restrict usage profiles to Codex history

8bbe6fd

vercel Bot deployed to Preview July 31, 2026 06:27 View deployment

HaodiFan mentioned this pull request Jul 31, 2026

fix: add local-state-first install preflight #48

Merged

40 tasks

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread scripts/evozeus-cli.mjs

fix: bind insights to approved history path

ff0756b

vercel Bot deployed to Preview July 31, 2026 09:43 View deployment

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread packages/runtime/src/evozeus_runtime/use_cases/run_codex_official_visualization.py

test: assert required source option structurally

3f58f2c

vercel Bot deployed to Preview July 31, 2026 09:53 View deployment

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread packages/runtime/src/evozeus_runtime/use_cases/run_codex_official_visualization.py

test: avoid direct click dependency

b213b8d

vercel Bot deployed to Preview July 31, 2026 10:02 View deployment

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread scripts/evozeus-cli.mjs

fix: scope insights to approved sessions

f08a61b

vercel Bot deployed to Preview July 31, 2026 12:19 View deployment

github-actions Bot added size:L Large PR and removed size:M Medium PR labels Jul 31, 2026

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread packages/runtime/src/evozeus_runtime/use_cases/run_codex_official_visualization.py

Comment thread packages/runtime/src/evozeus_runtime/use_cases/run_codex_official_visualization.py

fix(runtime): bind reports to approved session source

7ec3db7

vercel Bot deployed to Preview July 31, 2026 19:49 View deployment

HaodiFan mentioned this pull request Jul 31, 2026

feat: own UserPromptSubmit Lesson runtime #50

Merged

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread packages/runtime/src/evozeus_runtime/ledger/repository.py

fix(runtime): reject duplicate session identities

37003a5

vercel Bot deployed to Preview July 31, 2026 20:17 View deployment

chatgpt-codex-connector Bot reviewed Jul 31, 2026

View reviewed changes

Comment thread packages/runtime/src/evozeus_runtime/use_cases/scan_sessions.py

fix(runtime): validate approved sessions before state

527474f

vercel Bot deployed to Preview July 31, 2026 20:34 View deployment

HaodiFan requested review from Neillan96 and removed request for Neillan96 August 4, 2026 05:49

HaodiFan merged commit 6d1ccc0 into main Aug 4, 2026
16 checks passed

Uh oh!

Conversation

HaodiFan commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Linked Context

PR Scope

Instruction surface changed

Why this change is needed

Agent behavior before / after

Safety boundary

Prompt injection risk

EvoZeus Evidence Proof

Tests and Validation

Verdict / Review Gate

Risk Checklist

Privacy Checklist

Operational Checklist

Rollback plan

Versioning

AI-Assisted Work

Current Review State

Uh oh!

vercel Bot commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Jul 31, 2026

EvoZeus Privacy Scan

Uh oh!

github-actions Bot commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

EvoZeus Queue Guard

Uh oh!

github-actions Bot commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

EvoZeus Labeler Report

Uh oh!

github-actions Bot commented Jul 31, 2026

EvoZeus Dirty PR Check

Uh oh!

github-actions Bot commented Jul 31, 2026

EvoZeus Proof Gate

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

github-actions Bot commented Jul 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

EvoZeus Auto Response

Missing real behavior proof

Split this PR

Owner review required

Queue limit reached

Uh oh!

HaodiFan commented Jul 31, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

HaodiFan commented Jul 31, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

HaodiFan commented Jul 31, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

HaodiFan commented Jul 31, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

HaodiFan commented Jul 31, 2026 •

edited

Loading

vercel Bot commented Jul 31, 2026 •

edited

Loading

github-actions Bot commented Jul 31, 2026 •

edited

Loading

github-actions Bot commented Jul 31, 2026 •

edited

Loading

github-actions Bot commented Jul 31, 2026 •

edited

Loading