B · Tool permission + spend limit
The worker has permission and the amount is within the ceiling. B permits the request. In this illustration, execution carries the expired approval condition into an order effect.
diff --git a/.github/ISSUE_TEMPLATE/bug.yml b/.github/ISSUE_TEMPLATE/bug.yml index 2df691e..212b6f4 100644 --- a/.github/ISSUE_TEMPLATE/bug.yml +++ b/.github/ISSUE_TEMPLATE/bug.yml @@ -1,19 +1,36 @@ name: Bug report -description: Report a reproducible sandbox defect without sensitive data. +description: Report a reproducible app, documentation or research-implementation defect. body: - type: markdown attributes: - value: "For vulnerabilities follow SECURITY.md. Never post secrets, operational data or blind reference answers." + value: "For vulnerabilities follow SECURITY.md privately. Use synthetic examples; never post secrets, operational data or confidential reference answers. Review discussion exports before sharing them." + - type: dropdown + id: component + attributes: + label: Affected area + options: + - Guided browser experience + - Recorded evidence or source relationship + - Research engine, gateway or policy + - Documentation, build or repository + validations: + required: true - type: textarea id: reproduction attributes: label: Synthetic reproduction and environment - description: Include revision, OS, runtime versions and exact commands. + description: "For the app: route, case, inputs, ordered actions, browser/OS and viewport. For local code: commit, runtime versions and exact commands." validations: required: true - type: textarea id: behavior attributes: label: Expected and actual behavior + description: Explain what should happen and what you observed. Identify illustrative results versus recorded trials where relevant. validations: required: true + - type: textarea + id: evidence + attributes: + label: Safe supporting evidence + description: Optional screenshot, minimal log or source link. Redact sensitive content and keep the request or run identity if it explains the defect. diff --git a/.github/ISSUE_TEMPLATE/config.yml b/.github/ISSUE_TEMPLATE/config.yml index 3ba13e0..4740467 100644 --- a/.github/ISSUE_TEMPLATE/config.yml +++ b/.github/ISSUE_TEMPLATE/config.yml @@ -1 +1,8 @@ blank_issues_enabled: false +contact_links: + - name: APS user guide + url: https://github.com/SafetyMP/Agentic-Process-Safety/blob/main/docs/USER_GUIDE.md + about: Follow the comparison, decision path and reasoning workflow before opening a usage issue. + - name: Report a vulnerability privately + url: https://github.com/SafetyMP/Agentic-Process-Safety/security/advisories/new + about: Keep vulnerability details and sensitive evidence out of public issues. diff --git a/.github/ISSUE_TEMPLATE/proposal.yml b/.github/ISSUE_TEMPLATE/proposal.yml new file mode 100644 index 0000000..ef65e8a --- /dev/null +++ b/.github/ISSUE_TEMPLATE/proposal.yml @@ -0,0 +1,34 @@ +name: Research or design proposal +description: Propose a clearer investigation, case, control or evidence question. +body: + - type: markdown + attributes: + value: "Connect the proposal to APS's comparison → causal explanation → human reasoning experience. Proposed controls and new scenarios are not verified safeguards. Keep sensitive and blinded material private." + - type: textarea + id: problem + attributes: + label: Problem or research question + description: Who encounters the problem, in which case or workflow, and why does it matter? + validations: + required: true + - type: textarea + id: proposal + attributes: + label: Proposed change and alternative + description: What should change? Name a reasonable alternative or a competing explanation. + validations: + required: true + - type: textarea + id: evidence + attributes: + label: Supporting evidence and open questions + description: Link the actual source or observation. Distinguish modeled behavior, recorded trials, authored hypotheses and evidence still needed. + validations: + required: true + - type: textarea + id: acceptance + attributes: + label: How to assess the improvement + description: What observation would support the change, and what result would show it did not help? Separate implementation checks from user-learning or method-effectiveness claims. + validations: + required: true diff --git a/.github/ISSUE_TEMPLATE/question.yml b/.github/ISSUE_TEMPLATE/question.yml index 7345333..1392f4a 100644 --- a/.github/ISSUE_TEMPLATE/question.yml +++ b/.github/ISSUE_TEMPLATE/question.yml @@ -1,13 +1,18 @@ -name: Question or proposal -description: Ask a usage question or propose a focused improvement. +name: Usage or evidence question +description: Ask about the guided app, local development or what an observation establishes. body: - type: markdown attributes: - value: "Read README.md and SUPPORT.md. Keep vulnerabilities and confidential pilot evidence private." + value: "Start with README.md, docs/USER_GUIDE.md and SUPPORT.md. Keep vulnerabilities and confidential pilot evidence private." - type: textarea id: context attributes: - label: Question, context and proposed outcome - description: Include documentation consulted and a synthetic example. + label: Question and context + description: Include the documentation consulted and the case, route, command or evidence record you mean. Use a synthetic example. validations: required: true + - type: textarea + id: interpretation + attributes: + label: Your current interpretation + description: What do you think the result means, and which distinction or missing fact needs clarification? diff --git a/.github/PULL_REQUEST_TEMPLATE.md b/.github/PULL_REQUEST_TEMPLATE.md index e9fd349..07a7019 100644 --- a/.github/PULL_REQUEST_TEMPLATE.md +++ b/.github/PULL_REQUEST_TEMPLATE.md @@ -1,15 +1,18 @@ ## Problem and change -Describe the concrete problem, resulting behavior and scope. Link a sanitized issue where applicable. +Describe the concrete problem and resulting behavior. Identify the affected experience or research contract and link a sanitized issue where applicable. + +For an investigation change, explain what the user can now compare, inspect or decide. Distinguish a presentation change from a change to the model, recorded evidence or an operational control. ## Validation -List exact commands and actual results, including failures and missing prerequisites. Explain coverage. Include safe visual evidence for UI changes. +List exact commands and actual results, including failures and missing prerequisites. Explain coverage and independent review where applicable. Include safe visual evidence for UI changes; state which viewport, theme and interaction it shows. A local check does not establish publication or deployment. ## Review checklist - [ ] I have authority to contribute under Apache-2.0 and preserved third-party terms. - [ ] No secrets, operational/private data or blind reference answers are included. - [ ] Recorded evidence, human approvals and blinding/exposure records remain truthful. +- [ ] Retained comparison identity, reasoning and the separation of recommendation from authority remain intact where affected. - [ ] Dependencies, lockfiles, docs and tests are consistent; pinned gates remain intact. - [ ] Limitations and pending owner decisions are explicit; no production safety claim is implied. diff --git a/.github/workflows/repository-checks.yml b/.github/workflows/repository-checks.yml index 8d0f459..851d2f3 100644 --- a/.github/workflows/repository-checks.yml +++ b/.github/workflows/repository-checks.yml @@ -55,7 +55,7 @@ jobs: github.event.pull_request.base.ref == 'main' && github.event.pull_request.base.repo.full_name == github.repository && github.event.pull_request.head.repo.full_name == github.repository && - (github.event.pull_request.head.ref == 'release/aps-github-validation-v1' || github.event.pull_request.head.ref == 'codex/aps-redesign-release' || github.event.pull_request.head.ref == 'codex/aps-guided-investigation') + (github.event.pull_request.head.ref == 'release/aps-github-validation-v1' || github.event.pull_request.head.ref == 'codex/aps-redesign-release' || github.event.pull_request.head.ref == 'codex/aps-guided-investigation' || github.event.pull_request.head.ref == 'codex/aps-repository-alignment') run: node apps/aps-demo/scripts/qa-evidence.mjs - name: Record actual PR checkout and workflow identities id: validation @@ -81,7 +81,7 @@ jobs: github.event.pull_request.base.ref == 'main' && github.event.pull_request.base.repo.full_name == github.repository && github.event.pull_request.head.repo.full_name == github.repository && - (github.event.pull_request.head.ref == 'release/aps-github-validation-v1' || github.event.pull_request.head.ref == 'codex/aps-redesign-release' || github.event.pull_request.head.ref == 'codex/aps-guided-investigation') + (github.event.pull_request.head.ref == 'release/aps-github-validation-v1' || github.event.pull_request.head.ref == 'codex/aps-redesign-release' || github.event.pull_request.head.ref == 'codex/aps-guided-investigation' || github.event.pull_request.head.ref == 'codex/aps-repository-alignment') run: node apps/aps-demo/scripts/validation-evidence.mjs finalize - name: Retain predeployment evidence including failed runs if: >- @@ -89,7 +89,7 @@ jobs: github.event.pull_request.base.ref == 'main' && github.event.pull_request.base.repo.full_name == github.repository && github.event.pull_request.head.repo.full_name == github.repository && - (github.event.pull_request.head.ref == 'release/aps-github-validation-v1' || github.event.pull_request.head.ref == 'codex/aps-redesign-release' || github.event.pull_request.head.ref == 'codex/aps-guided-investigation') + (github.event.pull_request.head.ref == 'release/aps-github-validation-v1' || github.event.pull_request.head.ref == 'codex/aps-redesign-release' || github.event.pull_request.head.ref == 'codex/aps-guided-investigation' || github.event.pull_request.head.ref == 'codex/aps-repository-alignment') uses: actions/upload-artifact@cf430e030ddbb5b0abf93d22962f4752f3646cd9 # v7.0.2; ordinary artifact only with: name: aps-predeploy-${{ steps.validation.outputs.checkout_sha || github.sha }}-${{ github.run_id }}-${{ github.run_attempt }} diff --git a/AGENTS.md b/AGENTS.md new file mode 100644 index 0000000..6687900 --- /dev/null +++ b/AGENTS.md @@ -0,0 +1,46 @@ +# Working in APS + +APS is a research sandbox for investigating agent workflow failures through same-request comparisons, concrete causal paths, source-linked evidence and human reasoning. Read [README.md](README.md), [the vision](docs/VISION.md) and [the documentation map](docs/README.md) before substantial changes. The [user guide](docs/USER_GUIDE.md) describes the current app. Historical plans and receipts describe their own revisions; they are not current acceptance evidence. + +## Find the owning source + +| Concern | Source | +| --- | --- | +| App routes and shared investigation state | `apps/aps-demo/src/App.tsx`, `routes/`, `components/SessionContext.tsx` | +| Cases, immutable comparisons, saved reasoning and causal wording | `apps/aps-demo/src/domain/cases.ts`, `comparison.ts`, `session.ts`, `presentation.ts` | +| Fixed illustrative checks and retained legacy simulation | `apps/aps-demo/src/domain/simulation.ts` | +| Recorded public snapshots and source mapping | `apps/aps-demo/src/fixtures/`, `apps/aps-demo/src/data.ts` | +| HAZOP screening, gateway and procurement research | `packages/engine/`, `packages/governance/`, `packages/simulator/`, `policies/`, `schemas/`, `tests/` | +| Build/publication and evidence checks | `apps/aps-demo/scripts/`, `scripts/verify-linux-predeploy-evidence.mjs`, `scripts/verify-linux-ci-evidence.mjs`, `.github/workflows/` | + +Inspect the implementation and its focused tests before changing a contract. The React/TypeScript app uses Vite and one root npm lockfile. The separate Python/OPA research runtime is not a browser backend. Preserve these stacks unless the task explicitly changes them. + +## Invariants + +- Evaluate comparison lanes from the same immutable request and baseline. Input edits create a draft; they must not relabel retained results. Reasoning belongs to its selected run. +- Keep modeled outcomes, recorded trials, proposed hazards and human discussion separate. A recommendation or ready status grants no authority. A hypothetical recovery must retain the denied parent and name its starting state. +- Derive causal descriptions from the actual case and retained inputs. Passing approval freshness does not establish quotation validity, inventory truth or purchase necessity. An objective-conflict review is not objective arbitration. +- Preserve frozen fixture bytes, source provenance and historical receipts. The production gate protects the original research trees and exact approved additions. Do not alter a manifest, pin, assertion count or allowlist merely to make a failing check pass. +- Keep the published app static: no credentials, operational data, backend calls, telemetry or unreviewed external assets. Discussion exports can contain free text; do not publish them as test evidence without inspection. +- Preserve existing user work and generated evidence. Do not turn historical approvals, exposed reference material or previous successful runs into current authorization or independent validation. + +## Checks and delivery + +Run commands from the repository root. See [Contributing](CONTRIBUTING.md) for the change-to-check mapping and [development reference](docs/DEVELOPMENT_REFERENCE.md) for runtime prerequisites. + +```sh +npm ci +node scripts/verify-repository-readiness.mjs +node --test scripts/verify-repository-readiness.test.mjs +npm run demo:fixtures:check +npm run demo:production:check +npm run demo:lint +npm run demo:typecheck +npm run demo:test +npm run demo:build +npm run demo:static:check +``` + +Use `npm run demo:e2e` with the pinned Playwright browser for affected user journeys. Inspect the relevant desktop/mobile views, themes and keyboard focus; a screenshot does not prove interaction behavior. The root `npm test` invokes the Python gateway and requires the research setup, including OPA. Do not substitute a mock or skip an established gate to hide a missing prerequisite. + +For substantial changes, state scope and meaningful acceptance checks, obtain independent review, and report implemented, locally checked, CI-verified and deployed states separately. Match tests to changed behavior; preserve failure evidence. Refresh source-bound evidence after relevant changes rather than reusing an earlier receipt. Merging, deployment and external changes must stay within the user's authorized scope. diff --git a/CITATION.cff b/CITATION.cff new file mode 100644 index 0000000..a540d39 --- /dev/null +++ b/CITATION.cff @@ -0,0 +1,24 @@ +cff-version: 1.2.0 +message: "If you use APS in research or teaching, cite this repository and identify the commit or recorded evidence revision you used." +type: software +title: "Agentic Process Safety (APS)" +authors: + - name: "SafetyMP" + website: "https://github.com/SafetyMP" +repository-code: "https://github.com/SafetyMP/Agentic-Process-Safety" +url: "https://safetymp.github.io/Agentic-Process-Safety/" +license: Apache-2.0 +abstract: >- + An open-source research sandbox for investigating agent workflow hazards + through same-request control comparisons, case-specific causal paths, + recorded synthetic evidence and human reasoning. APS includes a static + guided investigation app and a separate HAZOP-inspired Python/OPA research + implementation. Illustrative outcomes and finite experiments do not + establish operational safety or the superiority of an analysis method. +keywords: + - agentic process safety + - multi-agent systems + - hazard analysis + - HAZOP + - agent governance + - human reasoning diff --git a/CODE_OF_CONDUCT.md b/CODE_OF_CONDUCT.md index 43db27e..4a3f40a 100644 --- a/CODE_OF_CONDUCT.md +++ b/CODE_OF_CONDUCT.md @@ -2,6 +2,6 @@ This policy applies to project issues, pull requests, reviews and project-managed spaces. Respect participants across backgrounds, identities and experience levels. Provide constructive feedback, protect privacy and give credit. Harassment, discrimination, threats, sexualized attention, doxxing and disclosure of confidential information are unacceptable. -For a public incident without sensitive details, use GitHub report-content controls or request maintainer review of the thread. For private abuse reports use https://support.github.com/contact/report-abuse. This is a GitHub platform route, not a project-operated mailbox. No private project conduct contact or dedicated enforcement team is established; creating one remains an owner task before community launch. +For a public incident without sensitive details, use GitHub report-content controls or request maintainer review of the thread. For private abuse reports use https://support.github.com/contact/report-abuse. This is a GitHub platform route, not a project-operated mailbox. No private project conduct contact or dedicated enforcement team is established; establishing a project-operated channel remains an owner task. The private vulnerability-reporting channel in [SECURITY.md](SECURITY.md) is for security reports, not a substitute conduct or appeals process. Maintainers with actual permissions may remove harmful content, request corrections, restrict participation or refer abuse to GitHub, considering severity, recurrence and participant safety. Affected participants may request reconsideration through an available maintainer channel without publishing private evidence. No confidential project appeal channel, independent enforcement capacity or response SLA is claimed. Do not retaliate against good-faith reports. diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index f7261de..0e6d93f 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -1,9 +1,43 @@ -# Contributing +# Contributing to APS -Start with [setup](README.md#setup-and-verification) and the [readiness audit](docs/open-source-readiness.md). This is a research sandbox with unverified hazards and no industrial safety approval. +APS helps people investigate when an agent's permitted action can still violate a workflow condition. Contributions should make the problem, comparison, causal explanation or human judgment clearer and better supported. Start with the [vision](docs/VISION.md), [user guide](docs/USER_GUIDE.md) and [documentation map](docs/README.md). The browser experience and Python/OPA research implementation have different contracts; identify which one your change affects. -Discuss substantial changes in a sanitized issue. Submit only work you have authority to license under [Apache-2.0](LICENSE); identify imported material and preserve third-party notices. No CLA or DCO process is established. Use [SECURITY.md](SECURITY.md) for vulnerabilities and [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md) for conduct concerns. +Useful contributions include a confusing step with a reproducible example, a missing evidence relationship, a case-specific explanation, an accessibility improvement, or a test that exposes a real contract violation. Use the repository's bug, question or research/design proposal forms. A proposal should state the user or research question, what the current evidence supports, and what observation would demonstrate an improvement. -Use focused branches and PRs describing the problem, behavior and actual validation. Use the root lockfile and `npm ci`. Run `node scripts/verify-repository-readiness.mjs`, `npm test`, `npm run typecheck` and frontend checks from the README. Python/policy changes require `bash scripts/verify.sh`; Phase 3 changes require `bash scripts/verify-phase3.sh` and its prerequisites. Report failures and missing prerequisites explicitly. +## Development setup -Keep pinned deployment and verification gates intact. Never replace recorded evidence with simulation. Do not post secrets, operational data or confidential reference answers. Preserve pilot blinding/exposure boundaries: exposed contributors cannot claim blinded independent review. Human assignments, approvals, ratings and signoffs remain pending until actual human work. Review is required before merge; response times are not guaranteed. +For the static app, use Node.js 22.18+ and the single root lockfile: + +```sh +npm ci +npm run demo:dev +``` + +The app needs no Python, OPA, account, API key or backend. For research-engine and gateway changes, use the separate [research setup](docs/DEVELOPMENT_REFERENCE.md#research-commands). Python 3.11+ and OPA on `PATH` are needed for the root simulator tests. Do not confuse the browser's illustrative checks with execution of those research controls. + +## Make and verify a focused change + +Use a branch and explain the concrete problem and resulting behavior in a pull request. Match checks to the change: + +| Change | Validation | +| --- | --- | +| Documentation or community files | `node scripts/verify-repository-readiness.mjs` and `node --test scripts/verify-repository-readiness.test.mjs`; inspect links and commands | +| Browser app | `npm run demo:lint`, `npm run demo:typecheck`, `npm run demo:test`, `npm run demo:build`, `npm run demo:static:check` | +| Browser interaction or presentation | The browser checks above, then `npm run demo:e2e` with the pinned Playwright browser; inspect relevant desktop/mobile screenshots | +| Frozen data or research boundaries | `npm run demo:fixtures:check`, `npm run demo:fixtures:audit`, `npm run demo:production:check`; explain any missing raw originals | +| Research engine, simulator or policy | Research prerequisites, `npm test`, `npm run typecheck`, `bash scripts/verify.sh`; consult [Phase 3 operations](docs/PHASE3_OPERATIONS.md) before `bash scripts/verify-phase3.sh` | +| Release evidence or workflows | Existing predeployment/postdeployment rejection tests and the applicable source-bound verification procedure | + +See the [development reference](docs/DEVELOPMENT_REFERENCE.md#current-development-workflow) for browser installation. Report exact commands, failures and missing prerequisites. A passing local test is not a deployment receipt. UI changes should include safe visual evidence and checks of focus, keyboard use, responsive layouts and both themes where affected. Obtain independent review before merging substantial changes. + +## Preserve the evidence contract + +- Keep browser-model outcomes, frozen recorded trials, authored hazards and human discussion distinct. New free-text reasoning or a ready status must not grant operational authority. +- Preserve immutable comparisons, their request identities and their attached reasoning. Editing inputs must not relabel an earlier result. +- Keep source pins, production-boundary manifests and release checks intact. A changed protected file needs a separately reviewed scope decision; updating a hash to hide a failure is not a fix. +- Retain historical receipts as observations of their original revision. Do not rewrite them to imply fresh acceptance or publish generated private artifacts. +- Preserve pilot blinding and exposure records. A contributor who saw reference answers cannot claim an independent blinded review. Human approvals, ratings and assignments require actual human work. + +## Rights and participation + +Submit only work you have authority to contribute under [Apache-2.0](LICENSE); identify imported material and preserve third-party terms. No CLA or DCO process is established. Never include secrets, operational/private data or confidential reference answers. Use [SECURITY.md](SECURITY.md) for vulnerabilities and [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md) for conduct concerns. Review and response times are not guaranteed. diff --git a/README.md b/README.md index bfd56e6..7c18fa6 100644 --- a/README.md +++ b/README.md @@ -1,113 +1,103 @@ -# Agentic Process Safety (APS): HAZOP for AI Agent Swarms +# Agentic Process Safety (APS) -Open-source research sandbox for **hazard and operability (HAZOP) analysis of AI agents and multi-agent systems**. APS combines deterministic hazard screening, a simulated procurement swarm, and policy-as-code experiments using **Open Policy Agent (OPA) and Rego**. +**Permission is only the first question.** APS is an open-source research and learning workspace for examining what must hold when an AI agent acts within a coordinated workflow: the right role, current approval, task-bound context, bounded delegation, and a defensible reason to proceed. -Compare the same illustrative request under different control bundles, follow the point where their outcomes diverge, and connect a proposed response to hazards and evidence. The separate research implementation models demand forecasting, inventory, supplier selection, finance approval and purchase execution. All tools are simulated; no real purchase or payment can be made. - -**SafetyMP research thesis — managing risk enables sustainable AI operations.** APS examines how delegation, shared context, and approval boundaries can introduce failure modes in multi-agent workflows. It links simulated hazards to proposed controls and repeatable experiments, asking whether risk can be managed without making agent collaboration impractical. These are research hypotheses, not proven production safeguards. See the [portfolio thesis](https://github.com/SafetyMP/SafetyMP). +[**Explore the live app**](https://safetymp.github.io/Agentic-Process-Safety/) · [Start the guided investigation](docs/USER_GUIDE.md) · [Project vision](docs/VISION.md) · [Documentation](docs/README.md) [](https://github.com/SafetyMP/Agentic-Process-Safety/actions/workflows/repository-checks.yml) -[](https://github.com/SafetyMP/Agentic-Process-Safety/actions/workflows/deploy-pages.yml) +[](https://github.com/SafetyMP/Agentic-Process-Safety/actions/workflows/deploy-pages.yml) [](LICENSE) -[](package.json) -[](pyproject.toml) - -> **Scope:** Research sandbox. Hazards remain unverified, requirements proposed, ratings null and actual human approvals zero. APS provides no safety certification or production protection. - -**Jump to:** [Demo](#demo) · [Quick start](#quick-start) · [Architecture](#architecture) · [Verification](#verification) · [Security](#security) · [Documentation](#documentation-map) · [Contributing](CONTRIBUTING.md) - ---- -## What you can study +## The problem APS makes tangible -- **AI agent risk analysis:** trace tool use, delegation, shared context and approval hazards through the [HAZOP methodology](docs/METHODOLOGY.md). -- **Agent governance and policy enforcement:** inspect the Python gateway and OPA/Rego controls in the [governance architecture](docs/GOVERNANCE_ARCHITECTURE.md), separate from browser-only simulations. -- **Reproducible multi-agent experiments:** compare baseline, conventional access checks and HAZOP-informed checks with [Phase 3 experiments and assurance](docs/PHASE3_OPERATIONS.md). Finite sandbox results do not establish real-world safety. +A purchase worker has permission to use its tool and stays within budget. Finance approved the purchase, but its request reaches the execution boundary **61 logical seconds later**, after the modeled approval has expired. One control bundle permits the request; another holds it. The purchase is still outstanding. -## Demo +APS connects that concrete situation to the next human question: **what evidence or authority would justify a different action?** Compare the same request, inspect the causal path, change one condition, and form a position that preserves supporting evidence, uncertainty and disagreement. -The React/TypeScript explorer runs entirely in the browser, with no account, API key, paid inference, backend or telemetry. +The project applies hazard and operability (**HAZOP**) analysis to AI agents and multi-agent systems. Its procurement setting makes dependencies between demand, inventory, supplier selection, finance and purchase visible. It is useful for engineers examining agent controls, researchers inspecting the method and evidence, and reviewers discussing what remains unresolved. -The opening case asks: **the purchase agent has permission, but is its approval still valid?** An illustrative purchase arrives 61 logical seconds after finance approval; the modeled validity rule requires an age below 60. The amount is 300 sandbox units, a teaching input rather than a recovered historical trial amount. +## Try the experience -The five views support one connected investigation: +The [live app](https://safetymp.github.io/Agentic-Process-Safety/) needs no account, API key, paid inference, installation or backend. -- **Start here:** launch the approval-expiry case, browse all nine illustrative cases, and inspect the declared procurement topology. -- **Case Lab:** compare B (tool permission and spend limit) with C (additional lifecycle checks) against identical inputs; optionally include A (no modeled checks). Inspect decisions, a selectable failure path, and bounded illustrative effects. -- **Hazard study:** examine the selected case's intended behavior, deviations, assumptions and proposed controls, with the full 96-entry screening register, node × deviation matrix and 16 authored references available. -- **Controls:** inspect what a protection checks, where its authority comes from, and its implementation/test mapping. The Case Lab is the single browser request runner. -- **Evidence:** inspect the retained browser comparison separately from recorded experiments, paired B/C calls, replay, charts, source lineage and explicit gaps. +1. Open the approval case and **Run comparison**. B checks tool permission and spend; C adds lifecycle conditions. Both receive the same request and starting state. +2. Follow the decision path, then choose **Try 59 seconds** and run again. The original comparison remains available. C now permits the request because the approval is current. +3. Choose **Form a position**. A recommendation opens a relevant evidence request, review question or eligible hypothetical recovery. Attach the modeled check or a related recorded event, explain the rationale, and state what would change your mind. -Choose **Run comparison**, then vary approval age using 59/60/61 shortcuts or the age input. At 59 both B and C allow; at 60 and 61 B allows and C denies. Changing inputs creates a draft and does not relabel an earlier result. Retained comparisons preserve their inputs, predictions and reasoning across views. Normal and unauthorized-tool cases also show where B and C reach the same outcome. Six secondary cases retain rule exploration with business effects explicitly unmodeled. +Five connected views support that investigation: **Start here → Case Lab**, with **Hazard study**, **Controls** and **Evidence** available as supporting inspections. The [user guide](docs/USER_GUIDE.md) covers all nine cases, saved discussions, exports and the distinction between a new request and a retained result. -The **Discussion worksheet** attaches a recommendation, rationale, evidence, assumptions, unanswered question and dissent to a particular run. Marking a position ready grants no authority. **Explore a fresh-approval request** creates a hypothetical child from C's no-effect state with a new request identity; it preserves the denied parent and does not continue B's already executed purchase. A fresh approval still does not establish that a quotation, forecast or stock record is correct. + -Use **Keep or export this discussion** for explicit browser-local save/load/delete, a readable Markdown brief, session JSON, and a configuration-only link. Notes remain in memory unless explicitly saved; loading a browser copy is an explicit action. Configuration links carry bounded inputs, not notes, outcomes or approvals. See the [redesign plan](docs/APS_EXPERIENCE_REDESIGN_PLAN.md) and [local implementation and verification record](docs/APS_REDESIGN_IMPLEMENTATION.md) for delivered scope and verification limits. +The guided experience was deployed on **2026-11-14** through [PR #32](https://github.com/SafetyMP/Agentic-Process-Safety/pull/32), at commit [`3ddc4ae`](https://github.com/SafetyMP/Agentic-Process-Safety/commit/3ddc4ae9a0dc42c5af98fa7818121cf18a1b364f). Its [deployment workflow](https://github.com/SafetyMP/Agentic-Process-Safety/actions/runs/43645203614) and 233-check browser run passed. This is a dated implementation result; badges and later runs describe subsequent revisions. -Keyboard entity buttons accompany the topology, the failure path has an ordered text representation, and charts include data tables. System, light and dark themes persist locally. The comparison explains these selected control bundles; it does not establish that HAZOP uniquely discovers the checks or outperforms other analysis methods. +## What the evidence means -The earlier executive journey, **Prepare example**, **Advance to next decision**, and decision-brief JSON import described in [continuity and saved brief usage](docs/CONTINUITY_IMPLEMENTATION.md) belong to a previous interface. That document is retained as historical implementation evidence, not instructions for this redesign. +| Material | What it supports | What it does not establish | +| --- | --- | --- | +| Illustrative browser comparison | How the fixed model's checks respond to the selected request | A live gateway decision, real purchase or complete business outcome | +| Frozen recorded experiments and replay | The retained synthetic trial's actual recorded calls and outcomes | That an edited browser request was executed in that trial | +| HAZOP findings and proposed controls | Questions, possible deviations and control mappings to investigate | Verified hazards, approved requirements or validated risk ratings | +| Human discussion | A recommendation, rationale, sources, uncertainty and dissent | Operational authorization or independent verification of the cited evidence | -**Project URL:** [safetymp.github.io/Agentic-Process-Safety](https://safetymp.github.io/Agentic-Process-Safety/). This README describes the redesign source; publication and source-bound CI acceptance for this revision require separate verification. The [earlier publication handoff](docs/STATIC_DEMO_PUBLICATION_HANDOFF.md) is retained as historical evidence and does not establish deployment of this redesign. +All hazards remain unverified, requirements proposed and risk ratings unset in the published fixtures. No real purchase or payment is made. The app does not establish production safety, safety certification, or that HAZOP outperforms another review method. Conventional systems can implement the same lifecycle checks. See the [vision and research questions](docs/VISION.md), [methodology](docs/METHODOLOGY.md) and [threat model](docs/THREAT_MODEL.md). -## Quick start +## Run locally -For the browser demo, install Node.js **22.18+** and npm. Run from the repository root using the single root lockfile: +For the browser app, use **Node.js 22.18+** and npm from the repository root. There is one root lockfile. -```bash +```sh git clone https://github.com/SafetyMP/Agentic-Process-Safety.git cd Agentic-Process-Safety npm ci npm run demo:dev ``` -Open the local URL printed by Vite at `/`. For a production build and local preview: +Open the local URL printed by Vite. To inspect the production build locally: -```bash +```sh npm run demo:build npm run demo:preview ``` -The browser demo is separate from the Python/OPA research implementation. It does not call OPA or authorize real operations. - -### Setup and verification - -The research engine and gateway also require **Python 3.11+** and **OPA 1.x** on `PATH`: +The separate research implementation requires **Python 3.11+** and **OPA 1.x** on `PATH`: -```bash +```sh python3 -m venv .venv .venv/bin/pip install -r requirements.lock npm ci -mkdir -p artifacts/phase2 -bash scripts/verify.sh > artifacts/phase2/full-verification.log 2>&1 +npm run demo ``` -Run `npm run demo` for the gateway-backed procurement simulation. Generated reports, CSV exports and verification logs stay in the ignored `artifacts/` directory. See the [development reference](docs/DEVELOPMENT_REFERENCE.md#setup-and-verification) for analysis CLI commands. +The browser app never calls this Python/OPA gateway. See [Phase 2 operations](docs/PHASE2_OPERATIONS.md) for setup and policy workflows, and [Phase 3 operations](docs/PHASE3_OPERATIONS.md) for experiments, offline dependencies and evidence inspection. -## Architecture +## Repository map -| Layer | Location | Responsibility | -| --- | --- | --- | -| HAZOP screening — Phase 1 | `packages/engine/`, `schemas/`, `examples/` | Deterministic findings, cascade paths, schema validation, review workflow and exports | -| Procurement swarm | `packages/simulator/` | TypeScript orchestrator and five simulated workers | -| Independent gateway — Phase 2 | `packages/governance/`, `policies/v1/` | OPA/Rego decisions, scoped identities, atomic transaction controls and signed authorization receipts | -| Experiments and assurance — Phase 3 | `packages/governance/`, `scripts/verify-phase3.sh` | Finite adversarial A/B/C experiments, telemetry, reports and source-bound execution attestation | -| Static explorer | `apps/aps-demo/` | Sanitized frozen snapshots and a separate illustrative TypeScript browser model | - -The explorer keeps **Recorded experimental results**, **Replayed test evidence** and **Illustrative browser simulation** distinct. Browser outcomes never replace frozen trial results. - -Read the [Phase 1 architecture](docs/ARCHITECTURE.md), [gateway architecture](docs/GOVERNANCE_ARCHITECTURE.md) and [methodology](docs/METHODOLOGY.md) for component and trust boundaries. +```mermaid +flowchart LR + H[HAZOP screening] --> F[Unverified findings and proposed controls] + G[Python and OPA research gateway] --> E[Recorded synthetic experiments] + F --> S[Frozen public snapshots] + E --> S + S --> U[Guided browser investigation] + M[Separate TypeScript illustration] --> U + U --> D[Human discussion and open questions] +``` -## Verification +| Location | Responsibility | +| --- | --- | +| [`apps/aps-demo/`](apps/aps-demo/) | React/TypeScript guided workspace; [browser model](apps/aps-demo/src/domain/README.md) and [fixture provenance](apps/aps-demo/src/fixtures/README.md) | +| [`packages/engine/`](packages/engine/) · [`schemas/`](schemas/) · [`examples/`](examples/) | Deterministic HAZOP screening, review contracts and authored examples | +| [`packages/simulator/`](packages/simulator/) | Procurement orchestrator and simulated workers | +| [`packages/governance/`](packages/governance/) · [`policies/v1/`](policies/v1/) | Research gateway, OPA/Rego controls, experiments and assurance tooling | +| [`tests/`](tests/) · [`scripts/`](scripts/) | Regression tests, integrity checks and evidence verification | +| [`docs/`](docs/README.md) | Current guides, technical references and clearly classified historical records | -Run the repository readiness check and frontend checks from the root: +## Verify a change -```bash +```sh node scripts/verify-repository-readiness.mjs -node --test scripts/verify-repository-readiness.test.mjs -npm run demo:fixtures:check -npm run demo:fixtures:audit npm run demo:production:check +npm run demo:fixtures:check npm run demo:lint npm run demo:typecheck npm run demo:test @@ -115,58 +105,12 @@ npm run demo:build npm run demo:static:check ``` -The fixture audit reports missing ignored originals explicitly and fails on differing available originals. For root simulator tests and types, complete the Python/OPA setup above, then run `npm test` and `npm run typecheck`. +Real-browser verification uses `npm run demo:e2e` with the pinned Playwright browser. Follow the browser setup in [Contributing](CONTRIBUTING.md) and the [development reference](docs/DEVELOPMENT_REFERENCE.md). Root `npm test` invokes the research gateway and needs the Python/OPA setup. Fresh acceptance must bind the exact source, workflow, artifact and reviewed screenshots; historical receipts cannot qualify new source. -Real-browser checks require the pinned Playwright browser. The runner defaults to Firefox on macOS and Chromium on Linux; unsupported values fail without fallback. See [browser setup](docs/DEVELOPMENT_REFERENCE.md#static-demonstration) for installation and `npm run demo:e2e`. +## Contribute and discuss -Phase 3 requires the pinned offline OpenTelemetry SDK before its complete assurance workflow: +Start with [Contributing](CONTRIBUTING.md), [Support](SUPPORT.md), the [Code of conduct](CODE_OF_CONDUCT.md), and [Security reporting](SECURITY.md). Useful contributions improve causal explanations, evidence mappings, reproducibility, accessibility and the quality of open questions. The [vision](docs/VISION.md) distinguishes current capability from research still needed. -```bash -.venv/bin/python -m pip install --no-index --find-links dependencies/otel-wheels --require-hashes -r dependencies/otel-wheels/requirements.lock -bash scripts/verify.sh -bash scripts/verify-otel-wheelhouse.sh -bash scripts/verify-phase3.sh -``` - -See [Phase 3 operations](docs/PHASE3_OPERATIONS.md) for fresh acceptance runs and evidence inspection. Historical logs and badges do not establish current browser acceptance, deployment success or human approval. - -## Security - -- Discussion notes and hypothetical fresh-approval branches affect local illustrative state only; they issue no operational authorization. The browser model omits authentication, real policy execution, atomic persistence, signed receipts and production objective arbitration. -- Local attestations, signing keys and ledger heads cannot independently establish identity or resist an owner replacing all local evidence. -- Finite experiment observations and unequal early-termination denominators are not real-world effectiveness estimates. Missing measurements remain unknown. -- Keep signing keys, databases and operational logs out of static deployments. Publish only reviewed public assets. -- Human review, risk acceptance and production decisions remain pending actual human work. - -Read the [threat model](docs/THREAT_MODEL.md), [security reporting policy](SECURITY.md) and [readiness audit](docs/open-source-readiness.md). Detailed experimental counts and evidence limitations remain in the [development reference](docs/DEVELOPMENT_REFERENCE.md#methodology-and-authenticity). - -## Community - -- [Contributing](CONTRIBUTING.md): setup, focused changes and required checks -- [Code of conduct](CODE_OF_CONDUCT.md): participation and enforcement expectations -- [Support](SUPPORT.md): questions and reproducible issue reports -- [Security](SECURITY.md): vulnerability reporting and supported scope +Coding agents should start with [AGENTS.md](AGENTS.md) for repository navigation and contribution constraints. Cite a specific revision using the [citation metadata](CITATION.cff); research results and interface revisions have different evidence boundaries. -Part of the [SafetyMP portfolio](https://github.com/SafetyMP). - -### Related projects - -- [FidusGate: MCP tool-call policy gateway](https://github.com/SafetyMP/FidusGate) explores policy decisions and signed receipts. -- [Corporate/Site Harness: evidence-gated software delivery](https://github.com/SafetyMP/corporate-site-harness) explores verification gates and human approval in agent-assisted development. - -## License - -Project-owned material is offered under [Apache 2.0](LICENSE), subject to preserved third-party terms and the pending source-data and rights review in the [readiness audit](docs/open-source-readiness.md). No release rights or industrial suitability are certified. - -## Documentation map - -| Topic | Start here | -| --- | --- | -| Analysis method and review gates | [Methodology](docs/METHODOLOGY.md) | -| Engine and gateway boundaries | [Phase 1 architecture](docs/ARCHITECTURE.md) · [Governance architecture](docs/GOVERNANCE_ARCHITECTURE.md) | -| Gateway operation and policy workflow | [Phase 2 operations](docs/PHASE2_OPERATIONS.md) | -| Experiments and attested assurance | [Phase 3 operations](docs/PHASE3_OPERATIONS.md) | -| Current browser redesign and local verification | [Redesign plan](docs/APS_EXPERIENCE_REDESIGN_PLAN.md) · [Implementation and verification](docs/APS_REDESIGN_IMPLEMENTATION.md) | -| Earlier browser QA and publication records (historical) | [Browser QA](docs/STATIC_DEMO_BROWSER_QA.md) · [Publication handoff](docs/STATIC_DEMO_PUBLICATION_HANDOFF.md) | -| Full CLI examples, human review syntax and retained evidence notes | [Development reference](docs/DEVELOPMENT_REFERENCE.md) | -| Open-source readiness and owner gates | [Readiness audit](docs/open-source-readiness.md) | +Part of the [SafetyMP portfolio](https://github.com/SafetyMP). Project-owned material is offered under [Apache-2.0](LICENSE), with third-party terms preserved. The [open-source readiness record](docs/open-source-readiness.md) documents remaining ownership and governance questions; a license is not certification of industrial suitability. diff --git a/SECURITY.md b/SECURITY.md index ef533e2..508da39 100644 --- a/SECURITY.md +++ b/SECURITY.md @@ -2,10 +2,16 @@ ## Supported versions -This pre-release prototype has no supported production release or LTS branch. The current development branch is the only candidate for best-effort investigation. Historical experiments, deployed snapshots and version 0.1.0 metadata are not maintenance commitments. No security response SLA is established. Do not use this sandbox as a production safety barrier. +APS is a research project with a public static investigation app and a separate local Python/OPA implementation. It has no supported production release or LTS branch. The current `main` branch is the candidate for best-effort investigation. Historical experiments, deployed snapshots and package-version metadata are not maintenance commitments. No security response SLA is established. Do not use this sandbox as a production safety barrier. + +## Scope and data + +The browser app makes illustrative comparisons; it does not call the Python gateway, authorize a purchase or submit discussion notes to a backend. Notes remain in memory unless a visitor explicitly saves a browser copy or downloads an export. An export may contain the visitor's free text; review it before sharing. The local research implementation has separate credentials, policy, persistence and receipt boundaries described in the [threat model](docs/THREAT_MODEL.md) and [governance architecture](docs/GOVERNANCE_ARCHITECTURE.md). + +Report incorrect security claims, unsafe static publication, exposed secrets, unintended network access, and violations of the declared gateway boundaries. An illustration's documented omission is not evidence that the omitted protection exists in production. ## Private reporting -Do not post vulnerability details, exploit payloads, keys or sensitive evidence in public issues or PRs. On the actual repository use Security → Advisories → Report a vulnerability **if enabled**. Enablement has not been verified here. No project security email or alternate confidential intake is verified. If the option is absent, a public issue may request only private reporting enablement, with no vulnerability details; retain the report privately until a confidential channel exists. +Do not post vulnerability details, exploit payloads, keys or sensitive evidence in public issues or PRs. Use GitHub's [private vulnerability report](https://github.com/SafetyMP/Agentic-Process-Safety/security/advisories/new), available under Security → Advisories → Report a vulnerability. Private vulnerability reporting was confirmed enabled through GitHub's authenticated API on 2026-11-14. This verifies the setting, not end-to-end intake, maintainer acknowledgment or a response SLA. No alternate confidential project intake is verified. If the option becomes unavailable, retain the report privately; a public issue may request only private-reporting enablement without vulnerability details. -In a private report include affected revision, environment, synthetic reproduction, impact and suggested mitigation. Never test live third-party services or purchasing systems. Maintainers should acknowledge privately, investigate, coordinate a fix/advisory and agree disclosure timing with the reporter. This is an intended process, not a staffed service. The owner must configure and test intake before public launch; see the [owner checklist](docs/open-source-readiness.md#pending-owner-checklist). +In a private report include affected revision, app or research component, environment, synthetic reproduction, impact and suggested mitigation. Never test live third-party services or purchasing systems. Maintainers should acknowledge privately, investigate, coordinate a fix/advisory and agree disclosure timing with the reporter. This is an intended process, not a staffed service. Testing confidential intake remains an owner task; see the [current owner decisions](docs/open-source-readiness.md#current-owner-decisions). diff --git a/SUPPORT.md b/SUPPORT.md index b53403b..c4d9a20 100644 --- a/SUPPORT.md +++ b/SUPPORT.md @@ -1,7 +1,18 @@ -# Support +# Getting help with APS -Start with the [README](README.md), [architecture](docs/ARCHITECTURE.md) and [readiness audit](docs/open-source-readiness.md) for setup, evidence limits and pending human review. See [CONTRIBUTING.md](CONTRIBUTING.md) for checks. +Try the [live app](https://safetymp.github.io/Agentic-Process-Safety/) and follow the [user guide](docs/USER_GUIDE.md) from comparison to a reasoned next step. The [vision](docs/VISION.md) explains the research question; the [documentation map](docs/README.md) separates current guidance from implementation references and historical records. For development, start with [Contributing](CONTRIBUTING.md). -On the actual repository, open a sanitized bug or question issue if Issues is enabled. Include revision, OS, Node/Python/OPA versions, commands, expected/actual behavior and a minimal synthetic reproduction. Remove secrets, personal information, operational data and confidential reference answers. Issue availability is unverified; there is no response SLA or commercial support commitment. Use [SECURITY.md](SECURITY.md) for vulnerabilities and [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md) for conduct concerns. +## Choose the right report -This project provides no industrial safety approval, risk acceptance or certification. Qualified humans must review evidence and authorize operational decisions outside this sandbox. +- **Bug:** a broken interaction, incorrect retained result, missing link or reproducible runtime failure. +- **Question:** using the app, understanding a case, running a command or interpreting an evidence boundary. +- **Research or design proposal:** a clearer investigation, a proposed case or control, a competing explanation, or evidence needed to assess a claim. +- **Vulnerability:** follow [Security](SECURITY.md) privately. Do not include vulnerability details in a public issue. + +Use the [issue chooser](https://github.com/SafetyMP/Agentic-Process-Safety/issues/new/choose). For an app report, include the route, selected case, input values, sequence of actions, browser/OS and viewport when relevant. State whether the observation concerns an illustrative comparison or a recorded trial. For a CLI report, include the commit, runtime versions, exact commands and a minimal synthetic reproduction. Remove secrets, personal information, operational data and confidential reference answers; review exported discussion notes before sharing them. + +## What support can establish + +Support can help explain modeled behavior and investigate defects. A discussion recommendation, successful test or sandbox trial does not establish that a real purchase should proceed, that a hazard has human approval, or that a system is safe. Operational decisions and risk acceptance remain outside this project. + +There is no response SLA or commercial support commitment. Use the [code of conduct](CODE_OF_CONDUCT.md) for participation concerns and the [current owner decisions](docs/open-source-readiness.md#current-owner-decisions) for unresolved maintenance and review responsibilities. diff --git a/apps/aps-demo/index.html b/apps/aps-demo/index.html index f1e81c4..f691cf5 100644 --- a/apps/aps-demo/index.html +++ b/apps/aps-demo/index.html @@ -4,51 +4,61 @@ -
APS is an open-source research sandbox applying hazard and operability (HAZOP) analysis to multi-agent AI systems.
-Follow five procurement workers and an orchestrator through demand forecasting, inventory, supplier selection, finance approval and simulated purchase execution.
-The interactive explorer includes a HAZOP Lab, simulated policy decisions, a deterministic swarm simulator and recorded experimental evidence. The research implementation uses Python, TypeScript and OPA/Rego.
+Permission is only the beginning. An AI agent can have the right tool and stay within budget while relying on an approval, task context, or earlier decision that no longer applies.
+APS is an open-source research sandbox for hazard and operability (HAZOP) analysis of coordinated AI agents. Compare the same request under different control bundles, follow the causal path, and form a human recommendation grounded in evidence.
+Finance approved a purchase, but the request arrives 61 logical seconds later. The model requires approval age below 60. Tool permission and spend checks permit it; an added freshness check stops it. Try 59 seconds, retain both results, and investigate what still needs to be known.
+The Case Lab connects nine illustrative cases to a hazard study, proposed controls and frozen recorded evidence. Recommendations lead to case-specific evidence questions, policy-review questions, or an explicitly hypothetical recovery. Your notes do not issue approval or change the controls.
+Read the project and evidence guide for a complete introduction without JavaScript, or use the interactive app when it loads. A plain-text source index is also available.
Hazards remain unverified and controls proposed. Browser simulation is separate from recorded experimental results. No real purchases, safety certification or production protection are provided.
-The interactive explorer requires JavaScript and runs entirely in your browser, without an account, API key, backend or telemetry.
+Hazards remain unverified and controls proposed. The illustrative browser model, frozen experiment records and separate Python/OPA research implementation have different evidence boundaries. No real purchase, payment, safety certification or production protection is provided.
+The interactive app runs entirely in your browser, without an account, API key, backend or telemetry. No live model inference occurs. Human learning and the value of the analysis method have not been established by these software checks.
AGENTIC PROCESS SAFETY · PROJECT GUIDE
+An agent can have the right tool and stay within budget while acting on an approval or task context that no longer applies. APS makes that problem concrete, then asks what the people responsible should do next.
+Agentic Process Safety is an open-source research sandbox applying hazard and operability (HAZOP) analysis to coordinated AI agent workflows. It connects intended behavior, possible deviations, proposed controls, evidence and human judgment. This guide works without JavaScript; the interactive app adds retained comparisons and a discussion worksheet.
+Demand, inventory and supplier workers contribute information. Finance supplies approval, purchase attempts the order, and an orchestrator coordinates the work. The final action depends on earlier conditions still holding.
+The worker has permission and the amount is within the ceiling. B permits the request. In this illustration, execution carries the expired approval condition into an order effect.
C also checks approval freshness and stops the request. No illustrative order is created under C. The purchase remains outstanding; a hold does not finish the task.
At 59 seconds both bundles permit the same request; at 60 and 61 seconds C stops it. Only approval age changes in the guided timing comparison. These are logical model values, not measured response times.
+B is the particular tool-and-spend comparator used here. Conventional systems can implement freshness and task-binding checks too. The comparison does not show that HAZOP uniquely discovers those checks or outperforms another review method.
+
Recommendations produce unresolved evidence requests, policy-review questions or, for the eligible denied expiry case, an explicitly hypothetical fresh-approval branch. Attach the selected modeled check or a related recorded event without replacing your notes. Assumptions and dissent remain visible in the discussion record.
+Notes stay in memory until explicitly saved in your browser. You can export a Markdown discussion brief or session JSON. A configuration link carries bounded inputs, not your notes or an approved outcome. Marking a position ready does not authorize an operation.
+The nine browser cases cover a currently authorized purchase, unauthorized tool use, excessive spending, recursive delegation, missing approval, replay, approval expiry, conflicting objectives and task-context mismatch. Six secondary cases do not project downstream business effects.
+Approval freshness is not quotation validity. The related recorded context mismatch occurs at demand, while the browser case evaluates a purchase request. The recorded goal-conflict case measures a budget violation, not objective arbitration. APS makes these relationships explicit before presenting the broader source records.
+For a first visit, open APS and follow the approval case. For deeper work, choose the source appropriate to your question:
+ +