aiwg 2026.8.28 → 2026.9.0

package/README.md

@@ -581,6 +581,7 @@ aiwg use forensics         # Digital forensics & incident response (13 agents, 1
 aiwg use marketing         # Marketing operations (37 agents, 87+ templates)
 aiwg use media-curator     # Media archive management (6 agents, 9 commands)
 aiwg use research          # Research workflow automation (8 agents, 8-stage pipeline)
+aiwg use civic-action      # Cited civic review and publication-preparation kit
 aiwg use rlm               # RLM addon (recursive context decomposition)
 aiwg use all               # Everything
 
@@ -635,8 +636,10 @@ The bundle is **byte-identical** in shape to its upstream form, so
 /plugin install compound-memory@aiwg
 ```
 
-The marketplace contains all 40 packaged frameworks and addons, so you can
-install only the capabilities a Claude Code workspace needs.
+The marketplace contains 40 independently packaged framework and addon
+plugins, so you can install only the capabilities a Claude Code workspace
+needs. Source-distributed opt-in addons such as Civic Action deploy with
+`aiwg use civic-action` and do not imply a marketplace wrapper.
 
 ### Multi-Platform Deployment
 
@@ -685,10 +688,11 @@ If you ship a product that bundles AIWG and want to be listed here, open an issu
 | **[Ops Complete](agentic/code/frameworks/ops-complete/)** | 12 | 3 | Operational infrastructure — incident management, runbooks, troubleshooting workflows |
 | **[Security Engineering](agentic/code/frameworks/security-engineering/)** | 2 | 5 | Applied security beyond STRIDE/OWASP — cryptographic primitive selection, chain-of-trust integrity, authentication-factor architecture, degraded-mode design, runtime secret hygiene, supply-chain trust, physical-access threats. Pattern-based, product-agnostic |
 
-### Addons (31)
+### Addons (32)
 
 | Addon | What It Does |
 |-------|--------------|
+| **[Civic Action](agentic/code/addons/civic-action/)** | Cited, opt-in decision support for source review, records planning, meeting reconciliation, public-technology research, local-resource profiles, corrections, and publication preparation |
 | **[RLM](agentic/code/addons/rlm/)** | Recursive context decomposition — process 10M+ tokens via sub-agent delegation with parallel fan-out |
 | **[Writing Quality](agentic/code/addons/writing-quality/)** | Content validation, AI pattern detection, authentic voice enforcement |
 | **[Testing Quality](agentic/code/addons/testing-quality/)** | TDD enforcement, mutation testing, flaky test detection and repair |

package/agentic/code/addons/civic-action/README.md

@@ -0,0 +1,53 @@
+# Civic Action
+
+Civic Action is an opt-in, vendor-neutral AIWG addon for evidence-bound civic
+research, public-meeting reconciliation, public-records planning, public
+technology review, local-resource indexing, corrections, and publication QA.
+
+It prepares structured artifacts and local findings. It never autonomously
+submits records requests, contacts people, identifies speakers, publishes
+content, or makes legal conclusions. Those transitions require named human
+review. A machine pass covers only the declared, schema-valid fields inspected
+by that command; it is not factual verification, compliance, or authorization.
+
+## Install and discover
+
+```bash
+aiwg use civic-action
+aiwg discover "plan a public records request"
+aiwg discover "reconcile a public meeting transcript with minutes"
+aiwg discover "publish cited local public information"
+```
+
+## Executable gates
+
+```bash
+aiwg civic source-gate source-registry.json
+aiwg civic meeting-gate vote-ledger.json reconciliation.json
+aiwg civic publish-gate publication-packet.json
+```
+
+All commands write a JSON report to stdout. Exit `0` means no blocking
+finding, exit `1` means one or more `block` findings, and exit `2` means invalid
+input or usage. Warnings remain in the report for human disposition.
+
+## Capability map
+
+| Need | Artifact |
+|---|---|
+| Public-source acquisition and reuse | `source-compliance-gate`, source registry schema/template |
+| Civic newsroom planning | `civic-newsroom-plan`, newsroom FlowPlaybook |
+| Meeting transcript, vote, and minutes | `public-meeting-reconcile`, vote/reconciliation schemas and flow |
+| Public-records preparation | `public-records-plan`, planning schema/template and manual-submission gate |
+| Public-technology evidence review | `public-technology-review`, neutral evidence/risk contract |
+| Newsroom personas and corrections | four bounded agents plus `editorial-correction-review` |
+| Publication QA | `civic-publish-gate`, gate-result schema and human publication gate |
+| Local public resources | `local-resource-index`, CAP/GTFS/HSDS profile contract |
+
+Optional frameworks add research, transcription, knowledge-base, editorial,
+operations, and security capabilities. If absent, skills emit an explicit
+`blocked-dependency-missing` or `manual-review-required` state; they do not
+claim those checks ran.
+
+See `docs/overview.md`, `docs/quickstart.md`, and the cited research reports in
+`docs/research/` for the design basis and limitations.

package/agentic/code/addons/civic-action/agents/citation-editor.md

@@ -0,0 +1,44 @@
+---
+name: citation-editor
+description: Validate civic claim-to-source alignment, provenance, freshness, and selectors without inventing evidence or treating citation presence as support.
+model: sonnet
+model-role: reasoning
+model-tier: standard
+tools: Read, Grep
+---
+
+# Citation Editor
+
+## Inputs
+
+- Required: versioned draft, material-claim map, source/retrieval registry, and selectors.
+- Optional: freshness policy, transcript sidecars, contrary evidence, and prior correction records.
+
+## Outputs
+
+- Per-claim gate results, evidence pointers, remediation, uncertainty, and overall citation status.
+
+## Responsibilities
+
+- Check that each material claim is entailed by its cited source version.
+- Verify source/retrieval IDs, hashes, page/section/cue/record selectors,
+  quoted text, freshness state, and claim epistemic status.
+- Reuse research-complete citation/provenance checks when installed.
+
+## Hard rules
+
+Never invent or repair evidence by changing the claim silently. Do not equate
+official provenance with truth, recency with correctness, or a URL with
+claim-level support. Missing dependencies or sources block the claimed check.
+
+## Output contract
+
+Return per-claim `pass|warn|block` results, evidence pointers, exact remediation,
+and an overall status. Preserve disputes for the human editor.
+
+## Recovery and scope
+
+Inspect only sources relevant to each claim. Independent claims may be checked
+in parallel. If evidence is missing, ambiguous, stale, or unreadable, record the
+failure and escalate to the editor; never retry by inventing or broadening the
+source set without authorization.

package/agentic/code/addons/civic-action/agents/civic-newsroom-operator.md

@@ -0,0 +1,57 @@
+---
+name: civic-newsroom-operator
+description: Orchestrate evidence-bound civic research and review packets without autonomously publishing, submitting requests, recording, or contacting people.
+model: sonnet
+model-role: reasoning
+model-tier: standard
+tools: Read, Grep, Task
+---
+
+# Civic Newsroom Operator
+
+## Inputs
+
+- Required: civic objective, jurisdiction context, permitted scope, and available source packet.
+- Optional: dependency inventory, prior gate results, correction history, and reviewer roster.
+
+## Outputs
+
+- A bounded workflow status, artifact/evidence map, gate results, unresolved uncertainty, and named human decisions still required.
+
+## Responsibilities
+
+- Resolve the requested workflow and assemble versioned source, jurisdiction,
+  claim, privacy, correction, and approval artifacts.
+- Verify source and jurisdiction assumptions before delegating or advancing.
+- Delegate only bounded research, citation, transcription, or review tasks.
+- Preserve dissent and blocked states when synthesizing specialist results.
+- Stop at every external-action and legal/editorial human gate.
+
+## Hard rules
+
+- Never publish, record, submit a request, contact a person, identify a speaker,
+  calculate an unreviewed legal deadline, or override a blocking gate.
+- Never turn an official record's allegation into a verified fact.
+- Never create a personal target dossier or operationalize harassment, doxxing,
+  intimidation, access-control bypass, or coordinated contact.
+- Missing optional capabilities produce `blocked-dependency-missing` or
+  `manual-review-required`, never fabricated results.
+
+## Output contract
+
+Return a workflow status, artifact paths and hashes, unresolved questions,
+machine-gate results, named human decisions still required, and safe next steps.
+Do not expose private chain-of-thought; report evidence and decision rationale.
+
+## Safety gates
+
+Apply `rules/civic-safety.md`, `rules/source-and-claim-integrity.md`, and
+`rules/publication-human-review.md`. A `block` is terminal for the proposed
+action. Non-overridable blocked classes have no routine exception path.
+
+## Recovery and scope
+
+Focus only on evidence relevant to the stated civic question. Independent
+source/citation reviews may run in parallel, but legal/editorial decisions may
+not. On missing, ambiguous, or failed evidence, preserve partial results,
+escalate to the human gate, and do not retry external action.

package/agentic/code/addons/civic-action/agents/correction-editor.md

@@ -0,0 +1,46 @@
+---
+name: correction-editor
+description: Review contested civic claims and prepare append-only correction records with retained provenance.
+model: sonnet
+model-role: reasoning
+model-tier: standard
+tools: Read, Grep
+---
+
+# Correction Editor
+
+## Inputs
+
+- Required: challenged artifact/hash, claim, complaint, original sources, and current publication state.
+- Optional: contrary evidence, response, retention holds, safety report, and downstream inventory.
+
+## Outputs
+
+- Schema-valid correction record, propagation checklist, uncertainty, and pending named-human decision.
+
+## Responsibilities
+
+- Compare the challenged claim, exact published version, source record,
+  contrary evidence, response, and correction policy.
+- Prepare a new version and visible correction note; never silently overwrite.
+- Track owned caches, feeds, indexes, exports, and syndication requests as
+  requested or observed outcomes, not guaranteed global erasure.
+
+## Hard rules
+
+Do not destroy evidence, reveal protected complainant data, retaliate, or approve
+your own correction. Preserve lawful holds and route privacy/safety urgency to a
+human. `withheld` and `canceled` items cannot re-enter publication automatically.
+
+## Output contract
+
+Return a correction record conforming to `correction-record.schema.json`, a
+downstream propagation checklist, unresolved retention/privacy questions, and a
+pending named-human decision.
+
+## Recovery and scope
+
+Focus on outputs derived from the challenged claim. Downstream inventories may
+be checked in parallel, but the correction decision is sequential and human.
+If evidence, retention authority, or safety impact is ambiguous, freeze silent
+changes, preserve the error state, and escalate without destructive retry.

package/agentic/code/addons/civic-action/agents/news-caster.md

@@ -0,0 +1,49 @@
+---
+name: news-caster
+description: Draft accessible civic explainers from approved evidence with claim-level citations and visible uncertainty.
+model: haiku
+model-role: efficiency
+model-tier: economy
+tools: Read, Grep
+---
+
+# News Caster
+
+## Inputs
+
+- Required: approved evidence packet, claim map, audience, scope, and style/accessibility constraints.
+- Optional: reviewed meeting packet, response record, correction history, and dissent notes.
+
+## Outputs
+
+- Draft, claim-to-source table, uncertainty/dissent section, accessibility checklist, and pending publication decision.
+
+## Responsibilities
+
+- Draft from supplied, versioned evidence only.
+- Attach a resolvable source selector to every material fact, number, date,
+  quote, vote, allegation, and consequential claim.
+- Distinguish `official_record`, `reported_allegation`, `verified_fact`,
+  `analysis`, `opinion`, and `unknown_or_disputed`.
+- Preserve uncertainty, contrary evidence, requests for response, accessibility
+  needs, and correction contact information.
+
+## Hard rules
+
+Do not browse for hidden personal information, infer speaker identity, invent
+citations, silently resolve conflicts, or claim publication approval. Treat
+machine transcripts and ledgers as drafts until human verified. Use
+institutional channels and public acts, not personal targeting.
+
+## Output contract
+
+Return a draft, claim-to-source table, uncertainty/dissent section,
+accessibility checklist, and `pending` human-publication decision. If the packet
+is incomplete, return `manual-review-required` with the missing evidence.
+
+## Recovery and scope
+
+Focus only on relevant approved evidence. Independent sections may be drafted in
+parallel only when their claim maps do not overlap. If a source, selector, or
+status is ambiguous, stop that claim, report the error, and escalate rather than
+guess or retry with unrelated context.

package/agentic/code/addons/civic-action/commands/civic-action.mjs

@@ -0,0 +1,45 @@
+import fs from 'node:fs/promises';
+import path from 'node:path';
+import { evaluateMeeting, evaluatePublication, evaluateSourceRegistry, exitCodeFor } from '../lib/gate-engine.mjs';
+import { assertCivicSchema, CivicSchemaValidationError } from '../lib/schema-validator.mjs';
+
+async function readJson(cwd, file) {
+  return JSON.parse(await fs.readFile(path.resolve(cwd, file), 'utf8'));
+}
+
+export default async function civicAction(args, context = {}) {
+  const cwd = context.cwd ?? process.cwd();
+  const subcommand = context.subcommand ?? args.shift();
+  if (args.includes('--help') || args.includes('-h')) return { exitCode: 0, message: 'Usage: aiwg civic source-gate <registry.json> | meeting-gate <ledger.json> <reconciliation.json> | publish-gate <packet.json>' };
+  try {
+    let result;
+    if (subcommand === 'source-gate') {
+      if (!args[0]) throw new Error('source-gate requires <registry.json>');
+      const registry = await readJson(cwd, args[0]);
+      assertCivicSchema('source-registry', registry);
+      result = evaluateSourceRegistry(registry);
+    } else if (subcommand === 'meeting-gate') {
+      if (!args[0] || !args[1]) throw new Error('meeting-gate requires <ledger.json> <reconciliation.json>');
+      const ledger = await readJson(cwd, args[0]);
+      const reconciliation = await readJson(cwd, args[1]);
+      assertCivicSchema('vote-ledger', ledger);
+      assertCivicSchema('meeting-reconciliation', reconciliation);
+      result = evaluateMeeting(ledger, reconciliation);
+    } else if (subcommand === 'publish-gate') {
+      if (!args[0]) throw new Error('publish-gate requires <packet.json>');
+      const packet = await readJson(cwd, args[0]);
+      assertCivicSchema('publication-packet', packet);
+      result = evaluatePublication(packet);
+    } else throw new Error(`Unknown civic gate: ${subcommand ?? '(missing)'}`);
+    assertCivicSchema('gate-result', result);
+    return { exitCode: exitCodeFor(result), message: JSON.stringify(result, null, 2) };
+  } catch (error) {
+    const payload = {
+      schema: 'aiwg.civic.gate-error.v1',
+      code: error instanceof CivicSchemaValidationError ? error.code : 'CIVIC_INPUT_INVALID',
+      error: error instanceof Error ? error.message : String(error),
+      ...(error instanceof CivicSchemaValidationError ? { validation_errors: error.validationErrors } : {}),
+    };
+    return { exitCode: 2, message: JSON.stringify(payload, null, 2) };
+  }
+}

package/agentic/code/addons/civic-action/docs/issue-traceability.md

@@ -0,0 +1,18 @@
+# Civic-action issue traceability
+
+| Issue | Implemented proof | Verification |
+|---|---|---|
+| #2213 legal/ethics guardrails | `rules/civic-safety.md`, `rules/publication-human-review.md`, jurisdiction template, cited legal brief, non-overridable source/publish gate findings | Source bypass and missing-review negative tests |
+| #2214 public-source compliance | Source registry and gate-result schemas/templates, explicit owner/method/cadence/decision/citation/fallback/empty-result fields, executable `source-compliance-gate` | Valid, hostile, public-record-alternative, empty-result, and last-good-copy tests; stable JSON/exit codes |
+| #2215 civic newsroom bundle | Non-core addon manifest, README, composition/degraded-mode map, `civic-newsroom-plan`, newsroom flow | Package and manifest discovery validation |
+| #2216 meeting transcript/votes | Vote-ledger and reconciliation schemas/templates with mover/seconder/timestamp/citation fields, executable meeting gate, meeting flow with two human gates | Conflict, inferred-vote, source-cue, required-field, exact-ID, and approval tests |
+| #2217 public records | Draft-only skill, full planning/tracker/response-ingestion schema and template, records flow, jurisdiction/manual-submission/privacy boundaries | Automatic-submission, tracking-field, schema-fixture, and Flow human-submission tests |
+| #2218 procurement | All requested source-class inventory, neutral evidence/risk/handoff schema, review template/skill, protected/public release states, no award recommendation | Inventory completeness and `award_recommendation: null` negative tests |
+| #2219 personas | Bounded operator, news-caster, citation-editor, and correction-editor agents; explicit skill/template allocation for resource/records/SEO-GEO work | Strict agent linter, model-policy, and placeholder/orphan tests |
+| #2220 publishing gates | Publication packet/gate schemas with section count/freshness, structured-data, link, correction, live-page, sitemap/reindex/cache and last-good-copy controls | Positive and multi-failure fixtures; JSON/exit-code tests |
+| #2221 local resources | CAP/GTFS/HSDS conditional field sets, citations, structured-data, template/skill, freshness/public-scope/correction/takedown and publish-gate handoff | Per-vertical required-field negative and research-source assertions |
+| #2222 packaging | Canonical opt-in addon, CLI namespace, docs, pilot/marketplace decision, optional dependency/degraded-mode contract; plugin deferred to thin generated wrapper | Dry-run and isolated provider deployment with live CLI execution |
+
+All machine passes remain evidence for review, not authorization for external
+action. The full readiness and packaging decision is in
+`docs/research/synthesis-and-readiness.md`.

package/agentic/code/addons/civic-action/docs/overview.md

@@ -0,0 +1,49 @@
+# Civic Action overview
+
+Civic Action packages research-first, evidence-preserving workflows for public
+sources, meetings, records requests, public-technology procurement, local
+resource indexes, corrections, and publication review. It is an opt-in addon,
+not a jurisdiction database, law practice, autonomous newsroom, recorder, CMS,
+or outreach system.
+
+## Safety and evidence model
+
+The executable source, meeting, and publication commands validate their input
+schemas and block selected declared conditions, including access-control bypass,
+unresolved source authorization/rights, conflicted or unverified vote evidence,
+uncited declared material claims, incomplete named privacy/accessibility review,
+and missing named exact-hash publication approval. Broader jurisdiction,
+consent, private/public status, anti-targeting, retention, and independent-review
+requirements are mandatory human workflow rules; the commands do not infer
+those facts from content.
+
+Source identity and retrieval versions are distinct. Contracts provide fields
+for epistemic status, source/retrieval identity, selectors, contrary evidence,
+and review history. Correction artifacts plan linked new versions rather than
+silent replacements; the addon does not enforce an external storage history.
+
+These choices follow [W3C PROV-O](https://www.w3.org/TR/prov-o/),
+[RFC 9309](https://www.rfc-editor.org/rfc/rfc9309.html), the
+[NIST AI RMF](https://www.nist.gov/itl/ai-risk-management-framework), the
+[NIST Privacy Framework](https://www.nist.gov/privacy-framework), and the
+[SPJ Code of Ethics](https://www.spj.org/spj-code-of-ethics/). The cited legal
+and standards research in `docs/research/` states scope and dissent explicitly.
+The `control-source-matrix.md` file maps each skill to its sources, tests, and
+executable-versus-declarative boundary.
+
+## Human authority
+
+Machine gates return `pass|warn|block`; a pass only means no blocking condition
+was found among the declared fields inspected by that command. It does not
+authorize action, determine legality, or independently verify facts. A named
+human reviews the exact artifact and underlying evidence before recording,
+submission, outreach, identity assignment, legal assertions, publication, or
+correction release. External action remains outside the executable gates.
+
+## Composition
+
+`aiwg-utils` supplies the Flow contract. Optional research-complete,
+media-curator, knowledge-base, media-marketing-kit, ops-complete, and
+security-engineering components add specialized capabilities. Their absence is
+reported honestly and never replaced with invented transcripts, citations,
+legal conclusions, accessibility conformance, or deployment evidence.

package/agentic/code/addons/civic-action/docs/quickstart.md

@@ -0,0 +1,45 @@
+# Civic Action quickstart
+
+> **First time using AIWG?** Begin with [Install, Connect, and Verify](https://docs.aiwg.io/pages/getting-started--install-connect-verify.html). This guide assumes AIWG is already installed and connected to the target project.
+
+Install the opt-in addon:
+
+```bash
+aiwg use civic-action
+```
+
+Start with a natural-language ask:
+
+```text
+Plan a public records request for human review.
+Reconcile this public meeting transcript with the approved minutes.
+Review this public technology procurement using source-linked evidence.
+Index local emergency, transit, and human-service resources with citations.
+Run civic publication quality gates on this evidence packet.
+```
+
+For source acquisition, copy `templates/source-registry.yaml`, replace every
+unknown state with observed evidence, obtain named review, then run:
+
+```bash
+aiwg civic source-gate source-registry.json
+```
+
+For meetings, preserve separate source media, transcript, ledger, and minutes
+artifacts. Do not replace `SPEAKER_XX` with a name without evidence and human
+confirmation:
+
+```bash
+aiwg civic meeting-gate vote-ledger.json reconciliation.json
+```
+
+For publication, supply claim-level citations, privacy and accessibility
+reviews, correction state, upstream gates, and exact-hash human approval:
+
+```bash
+aiwg civic publish-gate publication-packet.json
+```
+
+Exit `0` means no machine block, not permission to publish. Exit `1` is a
+blocking result. Exit `2` is invalid input/usage. Human publication remains a
+separate external action.