ai-core-framework 0.1.0

package/core/skills/meta-sync-platforms/SKILL.md

@@ -0,0 +1,53 @@
+---
+name: meta-sync-platforms
+description: Use when the user asks to run /sync-platforms, regenerate AI Core platform configuration, sync Claude/Cursor/Windsurf rules, or update generated agent command files.
+command: /sync-platforms
+display_name: "Sync Platforms"
+version: 1.0.0
+status: READY
+owner_agent: scrum-master
+model_preference: sonnet
+args: []
+preconditions:
+  - sync_script_exists: true
+  - ai_core_exists: true
+postconditions:
+  - claude_config_generated: true
+  - cursor_rules_generated: true
+  - windsurf_rules_generated: true
+---
+
+# /sync-platforms
+
+> Exports `core/` source-of-truth specs to Claude Code, Cursor, and Windsurf config files.
+
+## 🎯 Purpose
+
+Keep all supported AI tools aligned with the same agents, commands, and rules while editing only `core/` as the canonical source.
+
+## 🔄 Execution Flow
+
+1. Verify `core/scripts/sync-platforms.sh` exists.
+2. Run `bash core/scripts/sync-platforms.sh`.
+3. Review generated `.claude/`, `.cursor/rules/`, `.windsurfrules`, and `CLAUDE.md`.
+4. Confirm generated files should be committed.
+5. Report created/updated files.
+
+## 🔒 Hard Rules
+
+- Source changes MUST be made in `core/`, not generated exports.
+- MUST review generated files before commit.
+- MUST rerun after changing agents, commands, or rules.
+- MUST NOT delete user custom platform config without confirmation.
+
+## 📤 Outputs
+
+- Generated file list
+- Platform coverage
+- Suggested commit command
+- Next test steps for each tool
+
+## 🔗 Related Commands
+
+- `/setup-project`
+- `/validate-state`

package/core/skills/meta-ticket-health/SKILL.md

@@ -0,0 +1,61 @@
+---
+name: meta-ticket-health
+description: Use when the user asks to run /ticket-health, inspect ticket quality, find stale or blocked tickets, summarize ticket hygiene, or report health across AI Core tickets.
+command: /ticket-health
+display_name: "Ticket Health"
+version: 1.0.0
+status: READY
+owner_agent: scrum-master
+consults_agents:
+  - business-analyst
+  - tech-lead
+  - qa-tester
+model_preference: sonnet
+args: []
+preconditions:
+  - tickets_readable: true
+postconditions:
+  - health_report_generated: true
+---
+
+# /ticket-health
+
+> Finds stale, blocked, inconsistent, or under-specified tickets in `project/tickets/`.
+
+## 🎯 Purpose
+
+Keep large ticket sets healthy before backlog grooming, sprint planning, and release.
+
+## 🔄 Execution Flow
+
+1. Run `/validate-state`.
+2. Load tickets and backlog.
+3. Find tickets missing owner, estimate, AC, priority, or backlog rank.
+4. Find zombie tickets by state age.
+5. Find blocked tickets missing unblock plan.
+6. Find tickets not present in backlog.
+7. Find backlog items pointing to cancelled or done tickets.
+8. Output remediation actions.
+
+## 🔒 Hard Rules
+
+- MUST NOT silently change ticket state.
+- MUST distinguish errors from warnings.
+- MUST route remediation to correct owner agent.
+- MUST keep backlog order separate from ticket state.
+
+## 📤 Outputs
+
+- Critical errors
+- Warnings
+- Stale tickets
+- Blocked tickets
+- Backlog inconsistencies
+- Recommended handoffs
+
+## 🔗 Related Commands
+
+- `/validate-state`
+- `/backlog-status`
+- `/prioritize-backlog`
+- `/generate-views`

package/core/skills/meta-validate-audit-log/SKILL.md

@@ -0,0 +1,42 @@
+---
+name: meta-validate-audit-log
+description: Use when the user asks to run /validate-audit-log, verify AI Core audit-log integrity, check hash-chain continuity, or validate state history evidence.
+command: /validate-audit-log
+display_name: "Validate Audit Log"
+version: 1.0.0
+status: READY
+owner_agent: scrum-master
+requires_agents:
+  - scrum-master
+  - tech-lead
+model_preference: sonnet
+args: []
+preconditions:
+  - ai_state_exists: true
+postconditions:
+  - audit_log_hash_chain_checked: true
+---
+
+# /validate-audit-log
+
+> Validate the append-only audit log hash chain.
+
+## Purpose
+
+Detect manual edits to `project/audit-log.jsonl`. Executable workflow handlers append records with `prev_hash` and `hash`; this command recomputes the chain.
+
+## Chat Usage
+
+```text
+/validate-audit-log
+```
+
+The AI may call `core/scripts/validate-audit-log.sh` internally and must summarize the result.
+
+## Hard Rules
+
+### RULE VA-001: Audit lines are append-only
+Existing audit log lines MUST NOT be edited. Corrections require a new audit record.
+
+### RULE VA-002: Hash chain must validate
+Every line must point to the previous line hash and match its own payload hash.

package/core/skills/meta-validate-docs/SKILL.md

@@ -0,0 +1,58 @@
+---
+name: meta-validate-docs
+description: Use when the user asks to run /validate-docs, check documentation gates, validate changed docs against AI Core project rules, or verify docs before a PR, release, or handoff.
+command: /validate-docs
+display_name: "Validate Docs"
+version: 1.0.0
+status: READY
+owner_agent: tech-lead
+requires_agents:
+  - tech-lead
+consults_agents:
+  - qa-tester
+  - scrum-master
+model_preference: sonnet
+args:
+  - name: base
+    required: false
+    description: "Optional git base ref, e.g. origin/main"
+preconditions:
+  - git_repo_exists: true
+  - ai_state_exists: true
+postconditions:
+  - documentation_gates_checked: true
+---
+
+# /validate-docs
+
+> Run mechanical documentation and DoD evidence validation.
+
+## Purpose
+
+Catch missing docs before review or release. In chat, use `/validate-docs`; the AI may call `core/scripts/validate-docs.sh` internally and must summarize the result.
+
+## Execution Flow
+
+1. Identify changed files from staged changes, working tree, or the provided base ref.
+2. Check code changes have a ticket reference.
+3. Check API, setup, migration, architecture, and release-impacting changes have matching docs.
+4. Validate ticket `documentation.paths`, `adr.path`, `runbook.path`, and DONE-ticket DoD/QA evidence.
+5. Report blocking failures with exact remediation.
+
+## Chat Usage
+
+```text
+/validate-docs
+/validate-docs --base origin/main
+```
+
+## Hard Rules
+
+### RULE VD-001: Documentation is a gate
+If the script fails, the PR or state transition is not ready.
+
+### RULE VD-002: Ticket evidence must point to real files
+Declared docs, ADRs, runbooks, and QA evidence paths must exist.
+
+### RULE VD-003: DONE requires DoD fields
+DONE tickets must have machine-readable DoD fields and QA evidence.

package/core/skills/meta-validate-permissions/SKILL.md

@@ -0,0 +1,53 @@
+---
+name: meta-validate-permissions
+description: Use when the user asks to run /validate-permissions, check AI Core command permissions, validate agent ownership rules, or audit command access constraints.
+command: /validate-permissions
+display_name: "Validate Permissions"
+version: 1.0.0
+status: READY
+owner_agent: scrum-master
+requires_agents:
+  - scrum-master
+  - tech-lead
+model_preference: sonnet
+args: []
+preconditions:
+  - ai_state_exists: true
+  - command_specs_exist: true
+postconditions:
+  - agent_permissions_checked: true
+---
+
+# /validate-permissions
+
+> Validate that recorded state-history commands were executed by allowed agent roles.
+
+## Purpose
+
+Make role boundaries mechanically enforceable by comparing `project/tickets/*.json` history entries with `core/commands/**` command ownership metadata.
+
+## Chat Usage
+
+```text
+/validate-permissions
+```
+
+## Hard Rules
+
+### RULE VP-001: Command role must match
+`state_history[].by_agent` must match the command `owner_agent` or `requires_agents`.
+
+### RULE VP-002: Unknown commands fail
+Any `by_command` value without a matching `core/commands/**/<name>.md` file blocks validation.
+
+### RULE VP-003: No self-approval
+If `assignee` is recorded, the same agent cannot approve review-to-QA or DONE transitions.
+
+## Executable Runner
+
+The internal command runner also enforces command metadata before executing workflow handlers. Users normally type chat commands:
+
+```text
+/mark-ready TICKET-001
+/create-pr TICKET-001 https://github.com/org/repo/pull/1
+```

package/core/skills/meta-validate-state/SKILL.md

@@ -0,0 +1,58 @@
+---
+name: meta-validate-state
+description: Use when the user asks to run /validate-state, validate AI Core project state, check ticket or state consistency, or run state validator diagnostics.
+command: /validate-state
+display_name: "Validate State"
+version: 1.0.0
+status: READY
+owner_agent: scrum-master
+model_preference: sonnet
+args:
+  - name: ticket_id
+    required: false
+    format: "TICKET-XXX"
+preconditions:
+  - validator_script_exists: true
+postconditions:
+  - validation_report_generated: true
+---
+
+# /validate-state
+
+> Validates `project/` integrity, backlog references, ticket schema, and allowed state transitions.
+
+## 🎯 Purpose
+
+Catch corrupted backlog data, missing ticket references, skipped transitions, missing history, invalid schema, and stale workflow data before sprint planning, review, or release.
+
+## 🔄 Execution Flow
+
+1. Verify `jq` is installed.
+2. In chat, run `/validate-state` with optional ticket ID. The AI may call `core/scripts/validate-state.sh` internally.
+3. Validate `project/backlog/backlog.json` references existing tickets when validating all state.
+4. If available, run JSON schema validation against `core/config/ticket.schema.json` and `core/config/backlog.schema.json`.
+5. Report failures with file path, field, and recovery command.
+6. Block workflow if validation fails.
+
+## 🔒 Hard Rules
+
+- MUST NOT auto-rewrite state history.
+- MUST fail on invalid transitions.
+- MUST fail on current status not matching last history entry.
+- MUST warn on zombie tickets.
+- MUST fail when backlog references missing tickets.
+- MUST be run before `/plan-sprint` and `/release`.
+
+## 📤 Outputs
+
+- Pass/fail status
+- Ticket count
+- Backlog validation status
+- Failed ticket details
+- Suggested remediation
+
+## 🔗 Related Commands
+
+- `/mark-ready`
+- `/plan-sprint`
+- `/release`