@aicf/agent-skills 1.0.0-rc.4

package/.codex-plugin/plugin.json

@@ -0,0 +1,40 @@
+{
+  "name": "aicf-agent-skills",
+  "version": "1.0.0-rc.4",
+  "description": "Reusable Agent Skills for building, auditing, testing, migrating, and releasing AI Capability Framework projects.",
+  "author": {
+    "name": "AICF Contributors",
+    "url": "https://github.com/xennon101/ai-capability-framework"
+  },
+  "homepage": "https://github.com/xennon101/ai-capability-framework/tree/main/agent-skills",
+  "repository": "https://github.com/xennon101/ai-capability-framework",
+  "license": "MIT",
+  "keywords": [
+    "aicf",
+    "ai-capabilities",
+    "agents",
+    "agent-skills",
+    "evals",
+    "policy",
+    "governance",
+    "codex"
+  ],
+  "skills": "./skills/",
+  "interface": {
+    "displayName": "AICF Agent Skills",
+    "shortDescription": "Build governed, testable AICF capability systems with coding agents.",
+    "longDescription": "Agent Skills for authoring AICF capabilities, integrating runtime handlers, writing evals, red-teaming, validating provider adapters, and releasing public AICF projects safely.",
+    "developerName": "AICF Contributors",
+    "category": "Developer Tools",
+    "capabilities": ["Read", "Write"],
+    "websiteURL": "https://github.com/xennon101/ai-capability-framework",
+    "defaultPrompt": [
+      "Use AICF Agent Skills to author a new governed AI capability.",
+      "Use AICF Agent Skills to migrate this AI tool to AICF.",
+      "Use AICF Agent Skills to add evals and red-team tests for this capability."
+    ],
+    "brandColor": "#111827",
+    "composerIcon": "./assets/aicf-agent-skills-icon.svg",
+    "logo": "./assets/aicf-agent-skills-logo.svg"
+  }
+}

package/CHANGELOG.md

@@ -0,0 +1,28 @@
+# Changelog
+
+All notable public changes to the AICF Agent Skills package are documented here.
+
+## 1.0.0-rc.4 - Unreleased
+
+- Aligned the skills package, Codex plugin manifest, and all real skill metadata with
+  the AICF `1.0.0-rc.4` release candidate.
+- Added public publish metadata for `@aicf/agent-skills` and documented the
+  same-version, same-tag release process with the root package.
+- Added S6 release hardening: stale index checks, release-readiness assertions, stricter
+  package/install tests, and complete public package docs.
+- Added five S5 skills for provider conformance, optional storage/AWS adapters,
+  migration, docs/examples, and control-plane UI/API work.
+- Reached the full 17-skill public package inventory.
+- Added five S4 runtime and control skills for runtime integration, action lifecycle,
+  trust/redaction/retention, observability/replay, and controls/budgets.
+- Expanded trigger fixture coverage and package tests to cover 12 skills.
+- Added the first seven core AICF builder skills for capability authoring, governance
+  lifecycle, policy/risk, eval authoring, security red-team work, release hygiene, and
+  skills package maintenance.
+- Added trigger fixture coverage for all S3 skills and regenerated the public skill
+  index.
+- Added validation scripts, trigger coverage checks, public-safety checks, skill index
+  generation, installation tooling, package tests, and CLI dispatch for the skills
+  package.
+- Added the initial package skeleton, Codex plugin metadata, public docs placeholders,
+  and SVG assets.

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 AI Capability Framework contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,114 @@
+# AICF Agent Skills
+
+Reusable Agent Skills for coding agents that build, integrate, test, audit, and release
+AI Capability Framework projects.
+
+These are builder skills. They guide Codex or another compatible coding agent while it
+edits a repository. They are not AICF runtime capabilities, provider tools,
+application-agent tools, or a replacement for AICF validation.
+
+## What Is Included
+
+The package contains 17 public skills:
+
+- capability authoring, governance lifecycle, policy/risk, eval authoring, security
+  red-team work, release hygiene, and skill-pack maintenance;
+- runtime integration, action lifecycle, trust/redaction/retention,
+  observability/replay, and controls/budgets;
+- provider conformance, optional storage/AWS guidance, migration, docs/examples, and
+  control-plane UI/API guidance.
+
+See [docs/skill-index.md](docs/skill-index.md) for descriptions and trigger examples
+generated from the current package.
+
+## Install And Check
+
+Run package commands from this directory:
+
+```bash
+cd agent-skills
+npm install
+npm run check
+npm run pack:dry
+```
+
+The package is independent from the root AICF package. The repository root does not use
+npm workspaces for this package.
+
+## Commands
+
+```bash
+npm run validate
+npm run check:triggers
+npm run index
+npm run check:index
+npm run test
+npm run check:public
+npm run check:release
+npm run check
+npm run pack:dry
+node scripts/aicf-skills.mjs list
+```
+
+Expected successful output includes messages such as `Skill validation passed`,
+`Trigger coverage passed`, `Skill index is current`, and
+`Agent skills release readiness passed`.
+
+The root repository exposes equivalent script hooks such as `npm run skills:ci`,
+`npm run skills:check`, `npm run skills:pack:dry`, and `npm run skills:publish:dry`.
+
+## Local Skill Copy
+
+Copy the current skills to a local discovery folder:
+
+```bash
+node scripts/install-skills.mjs --target ../.agents/skills
+node scripts/install-skills.mjs --target $HOME/.agents/skills
+```
+
+The install script copies only skill folders. It does not copy package metadata, tests,
+plugin metadata, dependencies, or local artifacts. It refuses overwrite unless `--force`
+is passed.
+
+## Distribution
+
+The Codex plugin manifest lives at `.codex-plugin/plugin.json` and points to `./skills/`
+with relative asset paths. The package is released as `@aicf/agent-skills` and uses the
+same version and release tag as the root `ai-capability-framework` package.
+
+Prerelease install example:
+
+```bash
+npm install @aicf/agent-skills@next
+npx aicf-skills list
+```
+
+Before the first real scoped-package publish, the maintainer must configure npm
+ownership and Trusted Publishing for `@aicf/agent-skills`.
+
+Reference docs:
+
+- https://developers.openai.com/codex/skills
+- https://developers.openai.com/codex/plugins/build
+- https://agentskills.io/specification
+
+## Contributing
+
+New or changed skills must keep the required frontmatter, section order, one-level
+references/assets, trigger fixtures, and public-safety rules. Run:
+
+```bash
+npm run index
+npm run check
+```
+
+## Public Safety
+
+All skill content must be public-safe. Do not add private specs, local paths, prompt
+text captured from a real session, provider payloads, traces, credentials, account IDs,
+tenant IDs, or customer data. Use synthetic examples and `example.com` where examples
+need URLs.
+
+## License
+
+MIT, matching the root AI Capability Framework v1 license decision.

package/assets/aicf-agent-skills-icon.svg

@@ -0,0 +1,7 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 64 64" role="img" aria-labelledby="title desc">
+  <title id="title">AICF Agent Skills icon</title>
+  <desc id="desc">A simple public-safe icon for the AICF Agent Skills package.</desc>
+  <rect width="64" height="64" rx="12" fill="#111827"/>
+  <path d="M18 42V20h28v22H18z" fill="none" stroke="#f9fafb" stroke-width="4"/>
+  <path d="M24 28h16M24 36h10" stroke="#38bdf8" stroke-width="4" stroke-linecap="round"/>
+</svg>

package/assets/aicf-agent-skills-logo.svg

@@ -0,0 +1,9 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 420 96" role="img" aria-labelledby="title desc">
+  <title id="title">AICF Agent Skills logo</title>
+  <desc id="desc">Text logo for the AICF Agent Skills package.</desc>
+  <rect x="4" y="12" width="72" height="72" rx="14" fill="#111827"/>
+  <path d="M22 58V30h36v28H22z" fill="none" stroke="#f9fafb" stroke-width="5"/>
+  <path d="M30 39h20M30 49h12" stroke="#38bdf8" stroke-width="5" stroke-linecap="round"/>
+  <text x="96" y="42" font-family="Arial, sans-serif" font-size="28" font-weight="700" fill="#111827">AICF</text>
+  <text x="96" y="68" font-family="Arial, sans-serif" font-size="24" fill="#374151">Agent Skills</text>
+</svg>

package/docs/codex-plugin-distribution.md

@@ -0,0 +1,50 @@
+# Codex Plugin Distribution
+
+The Codex plugin manifest is `.codex-plugin/plugin.json`.
+
+The manifest points to `./skills/`, uses relative asset paths, and omits legal policy
+URLs unless the repository contains real public policy documents.
+
+The package does not include MCP servers, hooks, app connectors, or cloud deployment
+behavior. It distributes skill folders and public package metadata only.
+
+## Required Manifest Checks
+
+- `skills` is exactly `./skills/`.
+- `license` is `MIT`.
+- `version` matches `agent-skills/package.json`.
+- `interface.composerIcon` and `interface.logo` are relative asset paths that resolve
+  inside the package.
+- The manifest does not point to local-only files or dependency folders.
+
+## Local Marketplace Example
+
+Use a local marketplace entry while developing:
+
+```json
+{
+  "name": "aicf-local-plugins",
+  "plugins": [
+    {
+      "name": "aicf-agent-skills",
+      "source": {
+        "source": "local",
+        "path": "./agent-skills"
+      },
+      "policy": {
+        "installation": "AVAILABLE",
+        "authentication": "ON_INSTALL"
+      },
+      "category": "Developer Tools"
+    }
+  ]
+}
+```
+
+For public repository distribution:
+
+```bash
+codex plugin marketplace add xennon101/ai-capability-framework --sparse agent-skills
+```
+
+Run `npm run check:release` before sharing the plugin path.

package/docs/installation.md

@@ -0,0 +1,96 @@
+# Installation
+
+The AICF Agent Skills package is an independent nested npm package. Commands run from
+`agent-skills/` unless noted otherwise.
+
+## Local package setup
+
+```bash
+cd agent-skills
+npm install
+npm run check
+```
+
+Expected successful output includes `Skill validation passed`,
+`Trigger coverage passed`, and `Agent skills release readiness passed`.
+
+The root AICF repository can call package checks through root script hooks:
+
+```bash
+npm run skills:ci
+npm run skills:check
+npm run skills:pack:dry
+```
+
+## Repo-local skill copy
+
+Copy skills into a repository-local discovery folder:
+
+```bash
+node scripts/install-skills.mjs --target ../.agents/skills
+```
+
+Use `--force` only when replacing a previous local copy.
+
+## User-level skill copy
+
+Copy skills into a user-level discovery folder:
+
+```bash
+node scripts/install-skills.mjs --target $HOME/.agents/skills
+```
+
+The install script copies only skill folders. It does not copy package metadata, tests,
+plugin metadata, dependency folders, or local artifacts.
+
+## Codex plugin setup
+
+The plugin manifest is `.codex-plugin/plugin.json`. A local plugin marketplace entry can
+point at this package:
+
+```json
+{
+  "name": "aicf-local-plugins",
+  "interface": {
+    "displayName": "AICF Local Plugins"
+  },
+  "plugins": [
+    {
+      "name": "aicf-agent-skills",
+      "source": {
+        "source": "local",
+        "path": "./agent-skills"
+      },
+      "policy": {
+        "installation": "AVAILABLE",
+        "authentication": "ON_INSTALL"
+      },
+      "category": "Developer Tools"
+    }
+  ]
+}
+```
+
+For a public repository sparse install, use the real repository path:
+
+```bash
+codex plugin marketplace add xennon101/ai-capability-framework --sparse agent-skills
+```
+
+## npm package usage
+
+Consumers can install the npm package and use its CLI. Use the `next` dist tag for
+release candidates:
+
+```bash
+npm install @aicf/agent-skills@next
+npx aicf-skills list
+```
+
+The npm package does not install hooks or run provider calls. It ships public skill
+folders, docs, assets, and validation scripts.
+
+The skills package is released with the same version and Git tag as
+`ai-capability-framework`. Before the first real publish, the npm `@aicf/agent-skills`
+package must have scope ownership and Trusted Publishing configured for the repository
+publish workflow.

package/docs/public-release-checklist.md

@@ -0,0 +1,57 @@
+# Public Release Checklist
+
+Run these checks from `agent-skills/` before releasing or sharing the package:
+
+```bash
+npm run index
+npm run check
+npm run pack:dry
+node scripts/install-skills.mjs --target <temp-dir> --force --allow-outside
+```
+
+Run these checks from the repository root when available:
+
+```bash
+npm run skills:ci
+npm run skills:check
+npm run skills:pack:dry
+npm run skills:publish:dry
+npm run check:metadata
+npm run check:workspace-public
+npm run check:secrets
+```
+
+## Release Review
+
+- Confirm `docs/skill-index.md` is current.
+- Confirm `npm run check` passes, including `check:index` and `check:release`.
+- Inspect `npm run pack:dry` output and confirm it includes docs, assets, scripts,
+  plugin metadata, and all 17 skills.
+- Confirm tests, trigger coverage, and public-safety scans pass.
+- Confirm the plugin manifest at `.codex-plugin/plugin.json` points to `./skills/` and
+  uses relative asset paths.
+- Confirm `agent-skills/package.json`, `.codex-plugin/plugin.json`, and all real skill
+  metadata use the same version as the root package.
+- Confirm `agent-skills/package.json`, `.codex-plugin/plugin.json`, and all real skill
+  frontmatter use MIT, matching the root package and public license decision.
+- Confirm `agent-skills/package.json` is public and sets `publishConfig.access` to
+  `public`.
+- Confirm root release certification includes `skills:ci`, `skills:check`, and
+  `skills:pack:dry`.
+- Use `npm run skills:publish:dry` as a quick no-publish package check. For an actual
+  release-tag review, run the explicit tagged dry-run from the root release checklist,
+  such as `npm publish ./agent-skills --dry-run --access public --tag next`.
+- Confirm the install script copies skills to a temporary directory and does not copy
+  package metadata, tests, plugin metadata, dependency folders, or local artifacts.
+- Confirm the changelog, version consistency, and same-version, same-tag release policy
+  before tagging or publishing.
+- Confirm source archive check and npm package check behavior in the root repository
+  when preparing a full AICF release.
+- Confirm npm ownership and Trusted Publishing are configured for `@aicf/agent-skills`
+  before the first real scoped-package publish.
+
+## Public-Safety Review
+
+Do not include private specs, prompt text captured from a real session, provider
+payloads, traces, logs, credentials, account IDs, tenant IDs, generated docs, archives,
+dependency folders, or local-only files.

package/docs/skill-authoring-standard.md

@@ -0,0 +1,55 @@
+# Skill Authoring Standard
+
+Every AICF skill must be concise, public-safe, and focused on one builder workflow.
+
+## Folder And Frontmatter
+
+- Each skill folder lives directly under `skills/`.
+- The folder name and `SKILL.md` frontmatter `name` must match.
+- Names use lowercase letters, digits, and hyphens only.
+- Frontmatter must include `description`, `license: MIT`, `compatibility`, and the
+  required `metadata.aicf.skill.*` keys.
+- `metadata.aicf.skill.version` must match the package version.
+- `metadata.aicf.skill.package` must be `@aicf/agent-skills`.
+- `allowed-tools` is not permitted.
+
+## Required Sections
+
+Each `SKILL.md` must contain these required sections in order:
+
+1. `Purpose`
+2. `Use this skill when`
+3. `Do not use this skill when`
+4. `Inputs to inspect first`
+5. `Workflow`
+6. `Required outputs`
+7. `Validation`
+8. `Hard rules`
+9. `Handoff format`
+
+Keep `SKILL.md` under 500 lines. Put detailed checklists, templates, and examples in
+one-level resource files.
+
+## References, Assets, And Scripts
+
+- Links from `SKILL.md` may point only to one-level `references/`, `assets/`, or
+  `scripts/` files inside that skill folder.
+- Empty `references/` and `assets/` directories are not allowed.
+- JSON, YAML, and YML files must parse.
+- Do not create per-skill README, changelog, install guide, or extra process docs.
+
+## Trigger Fixtures
+
+Every production skill needs one trigger fixture in
+`tests/fixtures/trigger-prompts.json` with:
+
+- at least one positive example;
+- at least one negative example;
+- `required_description_terms` that appear in the skill description.
+
+## Public-Safe Content
+
+Skills must not copy private project notes, private planning files, local paths, real
+credentials, account identifiers, customer data, prompt text captured from a real
+session, provider payloads, traces, logs, generated docs, or archives. Use synthetic
+examples and `example.com`.

package/docs/skill-index.md

@@ -0,0 +1,157 @@
+# Skill Index
+
+Generated from the current AICF Agent Skills package.
+
+## aicf-action-lifecycle
+
+- Description: Implement or audit AICF propose, prepare, preview, approve, commit, verify, audit, idempotency, and rollback-safe action workflows for side-effecting AI capabilities.
+- Category: action-lifecycle
+- Reference files: references/action-state-machine.md, references/approval-patterns.md, references/idempotency-and-audit.md, references/verification-patterns.md
+- Primary trigger phrases: none
+- Positive examples: Implement prepare, approval, commit, and verify for a side-effecting capability. | Audit duplicate commit prevention and action state transitions. | Add idempotency and audit tests for approval-required actions.
+- Negative examples: Add a provider conformance export. | Create a quickstart README.
+
+## aicf-capability-authoring
+
+- Description: Author or modify AICF capability/entity manifests, schemas, model-facing slices, and starter evals. Use when creating an AI tool, exposing app functionality to models, or converting an operation into a governed AICF capability.
+- Category: capability-authoring
+- Reference files: references/checklist.md, references/examples.md, references/field-guide.md, references/manifest-decision-tree.md
+- Primary trigger phrases: creating an AI tool, exposing app functionality to models, converting an operation into a governed AICF capability.
+- Positive examples: Create an AICF manifest for a support ticket read tool. | Convert this refund operation into a governed AICF capability with an entity manifest. | Add starter evals for a new model-facing capability.
+- Negative examples: Implement the HTTP handler for an existing capability. | Prepare the repository for an npm release.
+
+## aicf-control-plane-ui
+
+- Description: Build or audit the optional AICF governance control-plane UI/API for capability catalogue, lifecycle, policy decisions, approvals, eval status, traces, controls, and evidence export.
+- Category: control-plane-ui
+- Reference files: references/api-contracts.md, references/approval-console.md, references/control-plane-screens.md, references/operational-ux.md
+- Primary trigger phrases: none
+- Positive examples: Build the AICF control-plane capability catalogue and approval queue. | Audit control-plane API responses for safe summaries only. | Add tests for lifecycle, eval status, controls, and evidence export views.
+- Negative examples: Convert a legacy tool into a capability manifest. | Validate Gemini function declarations.
+
+## aicf-controls-and-budgets
+
+- Description: Implement or audit AICF capability-level kill switches, circuit breakers, budgets, max loops, rate controls, and fail-safe read-only modes without replacing a model gateway.
+- Category: controls-budgets
+- Reference files: references/budget-policy.md, references/circuit-breakers.md, references/control-types.md
+- Primary trigger phrases: none
+- Positive examples: Add a capability kill switch and read-only mode. | Test budget exceeded behavior in the provider tool loop. | Audit circuit breakers for provider errors and validation failures.
+- Negative examples: Add lifecycle approval records. | Prepare a source archive review.
+
+## aicf-docs-and-examples
+
+- Description: Update AICF public docs, quickstarts, examples, README files, API references, provider guides, and sample apps so they match current public contracts and implementation.
+- Category: docs-examples
+- Reference files: references/docs-style-guide.md, references/example-app-checklist.md, references/public-api-docs.md, references/quickstart-template.md
+- Primary trigger phrases: none
+- Positive examples: Update the README and quickstart after adding a provider subpath. | Refresh example app docs so commands and expected output match implementation. | Add a provider guide with mock-first usage and optional live steps.
+- Negative examples: Implement a storage adapter. | Audit an approval state transition.
+
+## aicf-eval-authoring
+
+- Description: Create AICF capability-aware evals, golden tests, rubrics, deterministic fixtures, real-provider test cases, and regression suites. Use for AI quality, tool-choice accuracy, action correctness, or model upgrades.
+- Category: evals
+- Reference files: references/deterministic-scorers.md, references/eval-rubric.md, references/golden-test-patterns.md, references/real-provider-evals.md
+- Primary trigger phrases: none
+- Positive examples: Add golden evals for tool choice and input args. | Create deterministic fixtures for a model upgrade regression suite. | Write eval cases proving no forbidden commit tool is called.
+- Negative examples: Configure a kill switch for a capability. | Review npm package contents.
+
+## aicf-governance-lifecycle
+
+- Description: Manage AICF capability lifecycle, risk compiler rules, compatibility checks, and impact analysis. Use when moving capabilities through draft, review, approved, canary, production, deprecated, disabled, or removed states.
+- Category: governance
+- Reference files: references/compatibility-checks.md, references/impact-analysis.md, references/lifecycle-state-machine.md, references/riskcompiler-rules.md
+- Primary trigger phrases: moving capabilities through draft, review, approved, canary, production
+- Positive examples: Review whether this capability can move from draft to production. | Run an impact analysis before deprecating a capability. | Compare two capability versions for governance risk.
+- Negative examples: Write a new capability manifest from scratch. | Create trigger fixtures for the skills package.
+
+## aicf-migration
+
+- Description: Migrate existing AI tools, prompts, agents, raw function calls, RAG flows, or app integrations into AICF capabilities with policy, schemas, evals, runtime handlers, and safe rollout.
+- Category: migration
+- Reference files: references/conversion-playbook.md, references/legacy-tool-riskmap.md, references/migration-inventory.md, references/rollout-plan.md
+- Primary trigger phrases: none
+- Positive examples: Migrate our existing agent tools into AICF capabilities. | Inventory legacy function calls and convert them to governed manifests. | Plan a read-only canary rollout for a legacy RAG flow.
+- Negative examples: Build the optional control-plane UI. | Prepare an npm package dry run.
+
+## aicf-observability-replay
+
+- Description: Add AICF tracing, OpenTelemetry-style events, Langfuse/export adapters, replay fixtures, simulation, and trace-to-golden workflows while keeping raw prompts and provider payloads redacted by default.
+- Category: observability-replay
+- Reference files: references/redacted-telemetry.md, references/replay-simulation.md, references/trace-span-map.md, references/trace-to-golden.md
+- Primary trigger phrases: none
+- Positive examples: Add AICF trace events for routing, policy, tools, and eval scores. | Create a replay fixture from sanitized runtime behavior. | Draft a trace-to-golden eval with review required.
+- Negative examples: Wire host auth into the runtime context builder. | Create a red-team prompt injection case.
+
+## aicf-policy-and-risk
+
+- Description: Implement or audit AICF policy broker, risk tiers, semantic invariants, tenant isolation, permissions, entitlements, approval requirements, and fail-closed runtime decisions.
+- Category: policy-risk
+- Reference files: references/deny-approval-allow-matrix.md, references/policy-broker-contract.md, references/risktier-rules.md, references/semantic-invariants.md
+- Primary trigger phrases: none
+- Positive examples: Audit the policy broker for missing tenant checks. | Add fail-closed tests for approval-required prepare actions. | Review risk tiers and semantic invariants for a money movement capability.
+- Negative examples: Generate a provider conformance matrix. | Create public release notes.
+
+## aicf-provider-conformance
+
+- Description: Validate AICF capabilities across provider and framework adapters: OpenAI, Anthropic Claude, Google Gemini, Vercel AI SDK, MCP, LangChain/LangGraph, and Semantic Kernel-compatible workflows.
+- Category: provider-conformance
+- Reference files: references/provider-normalization-matrix.md, references/provider-runtime-boundaries.md, references/schema-compatibility.md, references/tool-call-mapping.md
+- Primary trigger phrases: none
+- Positive examples: Validate tool descriptors across OpenAI, Anthropic, Gemini, MCP, and LangChain. | Run a provider conformance matrix for routed support capabilities. | Check schema normalization and tool-call parsing across provider adapters.
+- Negative examples: Wire a handler registry into an API server. | Create public docs for a quickstart.
+
+## aicf-release-hygiene
+
+- Description: Prepare safe public AICF releases by checking package contents, source archives, npm dry runs, clean install, CI, docs, secrets, private artifacts, generated files, logs, traces, and provider payload leaks.
+- Category: release-hygiene
+- Reference files: references/forbidden-artifacts.md, references/package-smoke-test.md, references/release-checklist.md, references/source-archive-rules.md
+- Primary trigger phrases: none
+- Positive examples: Check package contents and clean install behavior before release. | Prepare a safe source archive review for the AICF package. | Run public release hygiene checks for docs, logs, traces, and generated files.
+- Negative examples: Write policy broker reason-code tests. | Create a new capability manifest.
+
+## aicf-runtime-integration
+
+- Description: Integrate AICF runtime into an application by wiring context builders, capability routers, handler registries, policy brokers, provider runtimes, stores, and tests. Use when connecting AICF to real app code.
+- Category: runtime
+- Reference files: references/capability-router-patterns.md, references/context-builder-patterns.md, references/handler-registry-patterns.md, references/policy-broker-adapter.md, references/runtime-integration-checklist.md
+- Primary trigger phrases: connecting AICF to real app code.
+- Positive examples: Wire AICF runtime into this API server with context builders and handlers. | Connect the capability router and policy broker to real app services. | Add runtime tests for denied permission, malformed args, and approval required.
+- Negative examples: Write only a capability manifest. | Prepare package contents for release.
+
+## aicf-security-redteam
+
+- Description: Generate AICF security and red-team tests for prompt injection, indirect injection, tool misuse, data exfiltration, cross-tenant access, approval bypass, schema confusion, and unsafe actions.
+- Category: security-redteam
+- Reference files: references/prompt-injection-patterns.md, references/redteam-acceptance.md, references/security-test-pack-map.md, references/tool-misuse-patterns.md
+- Primary trigger phrases: none
+- Positive examples: Generate red-team tests for approval bypass on a refund prepare capability. | Add prompt injection and schema confusion tests for a provider-exposed tool. | Create security cases for cross-tenant access and data exfiltration.
+- Negative examples: Author a normal read capability manifest. | Update the plugin metadata version.
+
+## aicf-skill-pack-maintenance
+
+- Description: Create, update, validate, test, and release the AICF Agent Skills Pack itself, including SKILL.md frontmatter, trigger descriptions, references, assets, plugin metadata, and public distribution.
+- Category: skill-pack-maintenance
+- Reference files: references/overlap-prevention.md, references/plugin-packaging.md, references/skill-authoring-rules.md, references/trigger-evaluation.md
+- Primary trigger phrases: none
+- Positive examples: Add a new SKILL.md with references and trigger descriptions. | Update the AICF Agent Skills plugin metadata. | Regenerate the skill index and validate trigger fixtures.
+- Negative examples: Promote a capability to production. | Create a red-team case for a high-risk tool.
+
+## aicf-storage-and-aws
+
+- Description: Implement optional AICF storage and AWS reference adapters for action stores, approval stores, audit, idempotency, Step Functions approvals, DynamoDB/RDS/Postgres, CloudWatch, and least-privilege IAM.
+- Category: storage-aws
+- Reference files: references/aws-adapter-boundaries.md, references/least-privilege-iam.md, references/step-functions-approval.md, references/store-interface-map.md
+- Primary trigger phrases: none
+- Positive examples: Add optional DynamoDB stores for AICF prepared actions and approvals. | Audit AWS adapter boundaries so root imports stay SDK-free. | Write fake-client tests for Step Functions approval handoff.
+- Negative examples: Run a provider conformance matrix. | Create an eval case for tool selection.
+
+## aicf-trust-redaction-retention
+
+- Description: Add AICF trust labels, taint tracking, provenance, provider-boundary redaction, trace redaction, and retention rules. Use for privacy, prompt injection, sensitive data, or logging boundaries.
+- Category: trust-redaction-retention
+- Reference files: references/redaction-rules.md, references/retention-rules.md, references/taint-rules.md, references/trust-labels.md
+- Primary trigger phrases: none
+- Positive examples: Add trust labels and taint rules for retrieved document context. | Review provider-boundary redaction for sensitive support data. | Add retention rules for trace and eval artifacts.
+- Negative examples: Implement a DynamoDB store. | Generate a skill index.
+