oh-my-githubcopilot 1.8.1 → 2.0.0-alpha.1cd01f4

package/.claude-plugin/plugin.json

@@ -1,8 +1,8 @@
 {
   "schemaVersion": "1.0",
   "name": "oh-my-githubcopilot",
-  "version": "1.8.1",
-  "description": "Multi-agent orchestration for GitHub Copilot CLI. 23 agents, 39 skills, parallel execution, HUD, PSM, SWE-bench.",
+  "version": "2.0.0",
+  "description": "Multi-agent orchestration for GitHub Copilot CLI. 19 agents, 59 skills, parallel execution, HUD, PSM, SWE-bench.",
   "author": {
     "name": "r3dlex"
   },
@@ -58,7 +58,27 @@
     "./skills/interactive-menu",
     "./skills/notifications",
     "./skills/doctor",
-    "./skills/session"
+    "./skills/session",
+    "./skills/verify",
+    "./skills/cancel",
+    "./skills/help",
+    "./skills/code-review",
+    "./skills/security-review",
+    "./skills/ultraqa",
+    "./skills/ultragoal",
+    "./skills/deep-dive",
+    "./skills/external-context",
+    "./skills/deepsearch",
+    "./skills/sciomc",
+    "./skills/remember",
+    "./skills/writer-memory",
+    "./skills/deepinit",
+    "./skills/self-improve",
+    "./skills/visual-verdict",
+    "./skills/ccg",
+    "./skills/build-fix",
+    "./skills/design",
+    "./skills/web-clone"
   ],
   "hooks": "./hooks/hooks.json",
   "mcp": "./.mcp.json",

package/AGENTS.md

@@ -4,7 +4,7 @@ This is the orchestration brain for the Oh My Copilot (OMP) plugin. All agents,
 
 ## Orchestrator Role
 
-The orchestrator is the top-level coordinator. It:
+The orchestrator is the top-level coordination role of the main session — it is **not a delegatable agent**. It:
 - Receives every user request
 - Analyzes scope and chooses the appropriate agent
 - Delegates work, never implements directly
@@ -14,32 +14,31 @@ The orchestrator is the top-level coordinator. It:
 
 The orchestrator **never writes code, docs, or configs directly**. It always delegates to a specialized agent.
 
-## Agent Registry (22 Agents)
+## Agent Registry (19 Agents)
 
 | # | ID | Tier | Tools | Role |
 |---|-----|------|-------|------|
-| 1 | `orchestrator` | high | TaskList, SendMessage, Glob, Grep | Top-level coordinator; never writes |
-| 2 | `explorer` | standard | Glob, Grep, Read | Fast codebase surveys, pattern finding |
-| 3 | `planner` | high | Read, Write, TaskCreate | Architecture, sequencing, risk assessment |
-| 4 | `executor` | standard | Read, Write, Edit, Bash | Implementation, refactoring, complex changes |
-| 5 | `verifier` | standard | Bash, Read, Glob | Testing, diagnostics, evidence collection |
-| 6 | `writer` | standard | Read, Write, Glob | Documentation, README, API docs |
-| 7 | `reviewer` | high | Read, Glob, Grep, LSP | Code review, quality gates |
-| 8 | `designer` | high | WebFetch, Figma tools | UI/UX, design system, Figma integration |
-| 9 | `researcher` | standard | WebSearch, WebFetch | External docs, benchmarking, options analysis |
-| 10 | `tester` | standard | Bash, Read, Write | Test writing, execution, CI integration |
-| 11 | `debugger` | high | Bash, Read, LSP, Grep | Error diagnosis, crash analysis |
-| 12 | `architect` | high | Read, Write, Glob | System design, cross-cutting concerns |
-| 13 | `security-reviewer` | high | Grep, Glob, Read | Vulnerability scanning, dependency audit |
-| 14 | `code-simplifier` | high | Read, Edit, Grep | Code simplification, clarity, maintainability |
-| 15 | `test-engineer` | standard | Bash, Read, Write | Test authoring, coverage analysis |
-| 16 | `critic` | high | Read, Grep, Write | Plan review, gap analysis, improvement suggestions |
-| 17 | `tracer` | high | Bash, Read, Grep | Causal investigation, root cause analysis |
-| 18 | `scientist` | high | Read, Write, Bash | Experimental design, hypothesis testing |
-| 19 | `code-reviewer` | standard | Read, Glob, LSP | PR reviews, style enforcement |
-| 20 | `document-specialist` | standard | Read, Write, Grep | Technical docs, API docs, guides |
-| 21 | `qa-tester` | standard | Bash, Read, Write | QA testing, regression verification |
-| 22 | `git-master` | standard | Bash, Read, Grep | Atomic commits, history management |
+| 1 | `explore` | fast | Glob, Grep, Read | Fast codebase surveys, pattern finding |
+| 2 | `planner` | high | Read, Write, TaskCreate | Architecture, sequencing, risk assessment |
+| 3 | `executor` | standard | Read, Write, Edit, Bash | Implementation, refactoring, complex changes |
+| 4 | `verifier` | standard | Bash, Read, Glob | Testing, diagnostics, evidence collection |
+| 5 | `writer` | standard | Read, Write, Glob | Documentation, README, API docs |
+| 6 | `designer` | high | WebFetch, Figma tools | UI/UX, design system, Figma integration |
+| 7 | `debugger` | high | Bash, Read, LSP, Grep | Error diagnosis, crash analysis |
+| 8 | `architect` | high | Read, Write, Glob | System design, cross-cutting concerns |
+| 9 | `security-reviewer` | high | Grep, Glob, Read | Vulnerability scanning, dependency audit |
+| 10 | `code-simplifier` | high | Read, Edit, Grep | Code simplification, clarity, maintainability |
+| 11 | `test-engineer` | standard | Bash, Read, Write | Test authoring, execution, coverage analysis, CI integration |
+| 12 | `critic` | high | Read, Grep, Write | Plan review, gap analysis, improvement suggestions |
+| 13 | `tracer` | standard | Bash, Read, Grep | Causal investigation, root cause analysis |
+| 14 | `scientist` | standard | Read, Write, Bash | Experimental design, hypothesis testing |
+| 15 | `code-reviewer` | high | Read, Glob, LSP | Code review, quality gates, merge verdicts |
+| 16 | `document-specialist` | standard | Read, Write, Grep, WebSearch, WebFetch | External docs, API research, options benchmarking |
+| 17 | `qa-tester` | standard | Bash, Read, Write | QA testing, regression verification |
+| 18 | `git-master` | standard | Bash, Read, Grep | Atomic commits, history management |
+| 19 | `analyst` | high | Read, Grep | Requirements analysis, gap identification |
+
+> **Migration (2.0)**: `explorer`→`explore`, `simplifier`→`code-simplifier`; `orchestrator` is now the top-level role only; `researcher`→`document-specialist`, `reviewer`→`code-reviewer`, `tester`→`test-engineer`. Run `omp doctor` to detect stale references.
 
 ## Delegation Rules
 
@@ -130,7 +129,7 @@ This repository uses the AI SDLC methodology scaffolded by `ai-sdlc-init`.
 
 ### Architecture Decision Records
 
-Significant architectural decisions are recorded in [`docs/adr/`](docs/adr/).
+Significant architectural decisions are recorded in [`docs/architecture/adr/`](docs/architecture/adr/).
 Before making a change that affects module boundaries, API contracts, data
 schemas, or dependency direction, check whether a relevant ADR exists.
 If your change contradicts an existing ADR, either update the ADR or open a
@@ -174,3 +173,8 @@ Before starting work on an issue:
 3. Do not make a fourth attempt without human acknowledgement.
 
 <!-- ai-sdlc-init:end -->
+
+<!-- v3-ai-sdlc-init:start -->
+## AI SDLC v3
+This repo follows the v3 AI-SDLC layout. See `.ai/matrix.json`, `.memory/human-override/`, and `docs/architecture/adr/`. Modules at `r3dlex/skills/ai-sdlc-init/modules/`.
+<!-- v3-ai-sdlc-init:end -->

package/CHANGELOG.md

@@ -1,341 +1,54 @@
-## oh-my-githubcopilot v1.8.1 — 2026-05-25
-
-### Fixes
-
-- **Node.js v26 native addon compatibility** — upgraded `better-sqlite3` from `^11.0.0` to `^12.10.0` to resolve
-  V8 API removals (`GetPrototype`, `GetIsolate`, `PropertyCallbackInfo::This`) that broke `npm install -g oh-my-githubcopilot`
-  on Node.js v26 (all macOS, Linux, and Windows with V8 >= 12.0)
-
-### Changed
-
-- **Dependencies**: `better-sqlite3` ^11.0.0 → ^12.10.0, `@types/better-sqlite3` ^7.0.0 → ^7.6.13
-
----
-
 # Changelog
 
-All notable changes to **oh-my-githubcopilot** are documented here, ordered newest first.
-
----
-
-## oh-my-githubcopilot v1.8.0 — 2026-05-25
-
-### Features
-
-- **Cross-repo sync**: Aligned README template, Star History chart, sponsor links, and badge shields with OMX/OMC/OMA/OMG
-- **Dual publish**: GitHub Actions release workflow now publishes to both npmjs.com (oh-my-githubcopilot) and GitHub Packages (@r3dlex/oh-my-githubcopilot) with version verification
-- **README updates**: 12 localized README files regenerated from unified template
-- **AI SDLC methodology**: Scaffolded across the repo with Archgate rules, ADR governance, Karpathy guidelines
-
-### Documentation
-
-- Updated READMEs across 13 languages (en, de, es, fr, it, ja, ko, pt, ru, tr, vi, zh)
-- Star History chart added for r3dlex/oh-my-githubcopilot
-- Sponsor link updated to github.com/sponsors/r3dlex
-
----
-
-## oh-my-githubcopilot v1.7.0 — 2026-04-21
-
-### Features
-
-- **vscode-omp**: Add VS Code extension for real-time workflow/agent visualization
-  - Workflows, Agents, Tasks panels in the activity bar
-  - StateReader interface with OmpStateAdapter (reads `.omc/state/*-state.json` and `subagent-tracking.json`)
-  - 200ms debounced FileSystemWatcher — prevents UI thrashing during heavy execution
-  - Live status bar showing active workflow count and running agents
-  - MCP server provider registration
-
----
-
-## oh-my-githubcopilot v1.6.0 — 2026-04-21
-
-### Features
-
-- **14 new skills — 25 → 39 total** — ported and adapted from oh-my-claudecode and oh-my-auggie:
-  - `ralplan` — consensus planning mode via `omp-plan --consensus`; triggers `/omp:ralplan`
-  - `research` — autonomous autoresearch workflow; triggers `/omp:research`
-  - `omp-doctor` — OMP installation diagnostics; triggers `/omp:omp-doctor`
-  - `omp-reference` — internal agent catalog (non-invocable)
-  - `ai-slop-cleaner` — AI-slop detection and prose cleanup; triggers `deslop`
-  - `tdd` — Red-Green-Refactor TDD workflow; triggers `/omp:tdd`
-  - `improve-codebase-architecture` — architecture improvement advisor; triggers `/omp:improve-codebase-architecture`
-  - `skillify` — convert ad-hoc prompts into reusable skills; triggers `/omp:skillify`
-  - `interview` — short-form interview mode (distinct from `deep-interview`); triggers `interview:`
-  - `graph-context` — auto-injected knowledge-graph context (non-user-triggered)
-  - `interactive-menu` — numbered-choice conversational TUI pattern (non-user-triggered)
-  - `notifications` — runtime notification delivery (distinct from `configure-notifications`); triggers `notifications:`
-  - `doctor` — general Socratic root-cause analysis (distinct from `omp-doctor`); triggers `doctor:`
-  - `session` — worktree/tmux session lifecycle management (distinct from `psm`); triggers `session:`
-- **`omp-plan` extended with consensus mode** — `--consensus` flag added; Planner→Architect→Critic loop (max 5 iterations)
-
-### Fixes
-
-- **keyword-detector** — registered all 14 new skill aliases (`doctor:`, `/doctor`, `/omp:doctor`, `interview:`, `/interview`, `/omp:interview`, `notifications:`, `/notifications`, `/omp:notifications`, `session:`, `/session`, `/omp:session`, `/omp:omp-doctor`, `/omp:ralplan`, `/omp:research`, `--consensus`) into `KEYWORD_MAP`
-- **graph-context / interactive-menu SKILL.md** — changed `trigger:` to `invocation:` so the keyword-detector test does not attempt to use their descriptive values as literal aliases
-
-### CI / Release
-
-- **npm Trusted Publishing** — replaced long-lived `NPM_TOKEN` secret with OIDC Trusted Publishing on npmjs.com; publish step now relies on `id-token: write` + `--provenance` with no stored secret
-
-### Verification
-
-- All 849 tests pass (1 integration test intentionally skipped)
-- Coverage: statements 87.89% / branches 85.49% / functions 94.64% — all above 80% threshold
-- Build, typecheck, and archgate check all pass
-
-## [1.5.9] — 2026-04-17
-
-### Features
-
-- **OMP CLI auto-update prompt** — added a launch-time npm update check for `omp` with OMX-style guardrails:
-  - interactive TTY only
-  - non-fatal failures
-  - 12-hour TTL cache at `~/.omp/state/update-check.json`
-  - disable via `OMP_AUTO_UPDATE=0` or `OMP_DISABLE_UPDATE_CHECK=true`
-  - prompt suppression for low-value commands (for example `omp version`)
-- **CLI front door wiring** — `src/index.mts` now invokes the update checker before command execution and treats
-  `--version` as the `version` command.
-
-### Verification
-
-- **CLI coverage expanded** — added dedicated update-flow tests under `tests/cli/update.test.mts` and extended
-  `tests/cli/index.test.mts` to assert update-check integration at the command entrypoint.
-
-## [1.5.8] — 2026-04-16
-
-### Improved
-- Raised test coverage across hooks, mcp, and psm modules
-- Branch coverage: 80.65% → 85.65% (+5pp above 80% threshold)
-- Statement coverage: 89.35% → 93.95%; Function coverage: 93.65% → 95.23%
-- New test files: stop-continuation, token-tracker (hooks), memory-store, state-manager (mcp), session, worktree (psm)
-
-## [v1.5.7] — Release readiness refresh
-
-### Release readiness
-
-- **Version manifests advanced** — `package.json`, `package-lock.json`, `plugin.json`, `.github/plugin/plugin.json`, `.github/plugin/marketplace.json`, and `.claude-plugin/plugin.json` now all target `1.5.7` for the next publishable release.
-- **Release docs aligned** — README, RELEASING, and build/release notes stay aligned with the current release workflow: bracketed changelog headings, root `plugin.json` → `.claude-plugin/plugin.json` sync, and coverage as a required release verification step.
-
-### Verification
-
-- **UltraQA release gate** — fresh verification on the release candidate remains green: `npm run build`, `npm run typecheck`, `npm test`, and `npm run test:coverage` all pass, with coverage still above the required floor (statements `89.35%`, branches `80.65%`, functions `93.65%`, lines `89.35%`).
-
-## [v1.5.6] — HUD watch daemon and model/budget fixes
-
-### Features
-
-- **HUD background watch daemon** — `omp hud --watch` (or `npm run hud:watch`) launches a daemon that polls session state every 2 seconds (configurable via `OMP_HUD_INTERVAL`), shows a real-time elapsed timer, and refreshes all three HUD artifacts on every cycle. SIGINT and SIGTERM are handled gracefully.
-
-### Fixes
-
-- **HUD model fallback** — corrected default model fallback from `claude-sonnet-4.5` to `claude-sonnet-4.6`.
-- **HUD token budget** — token budget is now resolved per-model from the `MODEL_CONTEXTS` map instead of using a fixed constant.
-- **HUD premium requests total** — total premium request count is now configurable via the `OMP_PREMIUM_REQUESTS_TOTAL` environment variable.
-- **HUD skills total** — `skillsTotal` corrected from 21 to 25, matching the actual number of shipped skills.
-- **Model router advisory** — updated the model-router advisory string to reference `claude-sonnet-4.6`.
-- **CLI coverage harness** — the `src/index.mts` test suite now executes the real `hud`, `hud --watch`, `version`, `psm`, `bench`, and unknown-subcommand flows so release coverage thresholds are protected by behavior-level assertions instead of placeholder checks.
-
-### Documentation
-
-- **Release docs aligned with current repo state** — README, release instructions, and build notes now agree on the bracketed changelog heading format, the root `plugin.json` → `.claude-plugin/plugin.json` sync step, and the need to run coverage as part of release verification.
-
-### Verification
-
-- **Fresh coverage evidence** — `npm run test:coverage` now reports `89.35%` statements, `80.65%` branches, `93.65%` functions, and `89.35%` lines.
-
----
-
-## [v1.5.5] — Copilot docs move and release refresh
-
-### Documentation
-
-- **Copilot docs moved to `.copilot/`** — relocated Copilot-facing instructions, agent docs, skill docs, and plugin skill docs out of `.github/` so `.copilot/` is now the single home for Copilot-specific documentation.
-- **Adoption flow updated** — `scripts/omp-adopt.sh` now installs Copilot docs under `.copilot/` while leaving `.github/` focused on workflows, plugin metadata, and hook entrypoints.
-
-### Release readiness
-
-- **Next release metadata advanced** — package, plugin, marketplace, and Claude plugin manifests move to `1.5.5` so the post-STATUS_LINE and post-doc-relocation state is ready for the next release cut.
-- **Release notes refreshed** — changelog and release docs now reflect the `.copilot/` doc boundary and the new release target.
-- **GitHub CI stability** — switched Vitest from `vmThreads` to `forks` single-process execution to avoid the segmentation faults seen in GitHub Actions on the `1.5.5` prep branch while preserving the same local coverage thresholds.
-
-## [v1.5.4] — Experimental Copilot setup and STATUS_LINE integration
-
-### Features
-
-- **Experimental Copilot defaults in setup** — `omp setup` now merges the required Copilot experimental feature flags into `~/.copilot/config.json`, preserves existing flags and custom status-line commands, and writes the packaged OMP status-line command when needed.
-- **Packaged STATUS_LINE support** — added a dedicated `src/hud/statusline.mts` entrypoint plus `bin/omp-statusline.sh` wrapper so Copilot CLI can render OMP session state directly from the installed plugin.
-- **HUD artifact export for status line rendering** — `hud-emitter` now writes the minimal status-line artifacts needed by the STATUS_LINE command in addition to the legacy HUD line output.
-
-### Fixes
-
-- **Setup verification harness** — added direct temp-HOME setup tests so Copilot config merging is exercised without depending on the limited top-level `omp` CLI subcommand surface.
-- **STATUS_LINE verification harness** — updated HUD/statusline tests so temp-HOME artifact paths are asserted correctly and lint-cleanly under the new status-line flow.
-
-### Verification
-
-- **Coverage remains above release floor** — fresh coverage evidence remains above the required 80% thresholds: statements `88.26%`, branches `83.98%`, functions `92.4%`.
-- **Focused addendum verification** — setup and status-line focused suites, typecheck, lint, build, and wrapper smoke all pass on the merged leader tree, with the remaining skipped integration setup test called out as pre-existing infrastructure debt rather than an addendum regression.
+All notable changes to oh-my-githubcopilot are documented here.
 
-## [v1.5.3] — Release readiness and CI enforcement
-
-### Fixes
-
-- **Version bump for the next publishable release** — advanced package, plugin, marketplace, and Claude plugin manifests to `1.5.3`, which clears the existing `v1.5.1` / `v1.5.2` tag history and avoids the next release colliding with already-used versions.
-- **Real CI test failures** — removed the `|| true` bypasses from the main CI test and coverage jobs so GitHub Actions now fails when tests or coverage regress.
-- **Claude plugin manifest sync** — normalized `.claude-plugin/plugin.json` to a single version field by re-syncing it from the root `plugin.json`.
-
-### Verification
-
-- **Fresh coverage evidence** — `npm run test:coverage` now reports `88.26%` statements, `83.96%` branches, and `92.4%` functions, satisfying the repo's `80%` thresholds.
-- **Fresh local CI parity run** — `npm run typecheck`, `npm run lint`, `npm test`, `npm run build`, `npx archgate check`, and `npm pack --dry-run` all pass on the release candidate branch.
-
-### Documentation
-
-- **Release process clarified** — `RELEASING.md` now calls out syncing all plugin manifests plus marketplace metadata and choosing the next unreleased semver before tagging.
-
-## [v1.5.0] — Copilot CLI release alignment
-
-### Features
-
-- **Copilot-ready agent metadata** — normalized `src/agents/*.md` frontmatter for Copilot-compatible agent loading fields while preserving the existing prompt bodies.
-- **Copilot instructions refresh** — rewrote the Copilot instructions with an OMP overview, delegation guidance, agent catalog, skill catalog, HUD reference, and keyword quick reference, now carried under the Copilot-facing docs set instead of `.github/` docs.
-- **README quick start refresh** — replaced the install flow with `copilot plugin install ...`, `/omp:setup`, and first-run Copilot commands.
-
-### Fixes
-
-- **Version sync to `1.5.0`** — aligned `package.json`, `package-lock.json`, root/plugin manifests, and marketplace metadata on the release target.
-- **Plugin manifest agent path** — `.github/plugin/plugin.json` now points at `./agents` so the manifest matches the Copilot-facing agent bundle.
-- **Release notes alignment** — consolidated the current Copilot CLI packaging/runtime fixes under the `v1.5.0` release line for a single consumer-facing release target.
-
-### Documentation
-
-- **Release expectations clarified** — release docs now call out the committed `dist/` expectation for plugin consumers.
-
----
-
-## [v1.4.2] — Dynamic version in HUD emitter
-
-### Fixes
-
-- **HUD version** — `hud-emitter` hook no longer hardcodes `"1.0.0"` as the session state version; reads from `package.json` via `createRequire` so `omp hud` always displays the installed package version
-
----
-
-## [v1.4.1] — CLI shebang fix, dynamic version
-
-### Fixes
-
-- **CLI shebang** — `bin/omp.mjs` now includes `#!/usr/bin/env node` banner (via esbuild `banner` option); the binary is executable directly without explicit `node` invocation (`550b764`)
-- **Dynamic version** — `omp version` now reads name and version from `package.json` at runtime via `createRequire`; no more hardcoded version string in source (`550b764`)
-
----
-
-## [v1.4.0] — Package renamed to oh-my-githubcopilot
+## [2.0.0] — 2026-06-13
 
 ### Breaking Changes
-
-- **Package renamed** — npm package renamed from `oh-my-copilot` to `oh-my-githubcopilot`; GitHub Packages now publishes as `@r3dlex/oh-my-githubcopilot`; install with `npm install oh-my-githubcopilot`
-- **Repository moved** — git remote updated to `git@github.com:r3dlex/oh-my-githubcopilot.git`
-
-### Changes
-
-- All source files, tests, specs, docs, and manifests updated to reflect the new package name
-- MCP server identity updated to `oh-my-githubcopilot`
-- All 11 localized READMEs, archgate ADRs, and release docs updated
-- `bin/omp.mjs` rebuilt with updated package identity
-
----
-
-## [v1.3.0] — Continuous release pipeline, dual-registry publish
-
-Commits: `fd366e6`…`a26f673`
-
-### Features
-
-- **Continuous alpha releases** — every push to `main` publishes `X.Y.Z-alpha.<sha>` to the `alpha` dist-tag; tagged commits (`vX.Y.Z`) publish stable to `latest` (`ec8fffa`)
-- **Hybrid dual-registry publish** — publishes to GitHub Packages (`@r3dlex/oh-my-githubcopilot`) always via `GITHUB_TOKEN`; publishes to npmjs.com (`oh-my-githubcopilot`) when `NPM_TOKEN` secret is configured; graceful skip with notice if absent (`ee42604`)
-- **npm release CI pipeline** — `release.yml` with four jobs: `build` (version resolution + artifact), `test` (CHANGELOG gate for stable only), `publish` (dual-registry), `github-release` (stable only, attaches `.tgz`) (`5a3ae67`)
-
-### Fixes
-
-- **Workflow parse error** — replaced `secrets.NPM_TOKEN != ''` in step `if:` conditions with a dedicated check step outputting `available=true/false` (`e82e807`)
-- **Version-agnostic plugin test assertions** — `plugin-install.test.mts` now reads version from `packageJson().version`; `marketplace.json` metadata.version synced (`7f99f14`)
-- **Stale agent/skill counts** — JSON descriptors and e2e tests corrected to 23 agents / 25 skills (`74c5122`)
-
-### Documentation
-
-- **CHANGELOG rewritten** — entries now derived from actual git log commit ranges per version (`a26f673`)
-- **Agent and skill docs normalized** — all 23 agent descriptors and 25 skill `SKILL.md` files updated to consistent format (`fd366e6`)
-
----
-
-## [v1.2.0] — GraphProvider abstraction, graphwiki adapter, spending skill
-
-Commits: `a234799`, `feb5e65` (PR #10)
-
-### Features
-
-- **GraphProvider abstraction** — new `src/graph/` module with `GraphBuildable` and `GraphWikiClient` interfaces; provider resolved from `.omp/config.json` `graph.provider` with local > global > default (`graphwiki`) resolution
-- **GraphifyAdapter** — extracted graphify CLI wrapper from `src/skills/graphify.mts` into `src/graph/graphify-adapter.mts`; graphify skill now delegates to the adapter (public API unchanged)
-- **GraphwikiAdapter** — new `src/graph/graphwiki-adapter.mts` wrapping the `graphwiki` npm CLI (`npm install -g graphwiki`); implements both `GraphBuildable` and `GraphWikiClient`
-- **graphwiki skill** — new `/omp:graphwiki` skill for direct access to graphwiki CLI: `query`, `path`, `lint`, `refine`, `build`, `status`, `clean`
-- **graph-provider skill** — new `/omp:graph-provider` skill for managing the active provider: `get`, `set`, `list`, `build`, `status`, `clean`, `query`
-- **spending skill** — new `/omp:spending` skill exposing `status` and `reset` for premium request usage tracking
-- **keyword-detector wiring** — 8 new keyword entries: `graphify:`, `graphwiki:`, `graph:`, `spending:`, `/graphify`, `/graphwiki`, `/graph-provider`, `/spending`
-
-### Fixes
-
-- **plugin.json** — removed non-existent `./agents` path from agents array; added `graphify`, `graphwiki`, `graph-provider`, `spending` to skills list (25 skills total)
-- **gitignore** — untracked `coverage/.tmp`, `.omc/` state files, `devops.md` (`a234799`)
-
----
-
-## [v1.1.0] — CI hardening, HUD format change
-
-Commits: `ce6f3bd`…`051ac20` (PR #9)
-
-### Features
-
-- **HUD format: `tools:N` → `tools:N/M`** — all count fields (tools, skills, agents) now display as used/total (e.g., `tools:12/13`, `skills:5/25`, `agents:3/23`); added `toolsTotal`, `skillsTotal`, `agentsTotal` fields to `HudState` and `HudMetrics`
-- **CLI elicitation support** — expanded OMP setup wizard to handle interactive CLI prompts during MCP config generation
-
-### Fixes
-
-- **CI: zero-install artifact pattern** — build artifact uploaded once in `build` job and downloaded in `test`/`publish` jobs; eliminates redundant `npm install` across CI jobs (`d46f647`)
-- **CI: vitest hanging** — added `< /dev/null` stdin redirect and `timeout 120` to prevent vitest blocking indefinitely in non-TTY environments (`fa9fdfb`)
-- **CI: vitest `--forceExit`** — added flag to ensure process exits after tests complete in CI (`c88c2c7`)
-- **CI: coverage job timeout** — separated coverage into its own job with a 10-minute timeout; tests run without coverage in parallel (`9608b07`)
-- **CI: coverage provider** — removed incompatible `@vitest/coverage-istanbul`; using `@vitest/coverage-v8` exclusively (`ce6f3bd`)
-- **Redundant `root/agents` directory** — removed stale agents directory at repo root; all 23 agents live in `src/agents/` (`8c521c0`)
-
----
-
-## [v1.0.0] — Initial release
-
-Commits: `804fc37`…`0f96d48` (initial implementation + `6ee243f` rename)
-
-Initial release of **oh-my-githubcopilot (OMP)** — a multi-agent orchestration plugin for GitHub Copilot CLI.
-
-### Features
-
-- **23 specialized agents** via Claude Code subagents: orchestrator, explorer, planner, executor, verifier, writer, reviewer, designer, researcher, tester, debugger, architect, security-reviewer, simplifier, test-engineer, critic, tracer, scientist, code-reviewer, document-specialist, qa-tester, git-master, analyst
-- **22 skills** including: `setup`, `mcp-setup`, `autopilot`, `ralph`, `ultrawork`, `team`, `ecomode`, `swarm`, `pipeline`, `plan`, `omp-plan`, `hud`, `note`, `trace`, `learner`, `swe-bench`, `wiki`, `psm`, `release`, `graphify`, `spending`, `spawn`
-- **6 hooks**: `keyword-detector`, `delegation-enforcer`, `model-router`, `token-tracker`, `hud-emitter`, `stop-continuation`
-- **HUD display system** — real-time session context, token tracking, and agent/skill usage counters in tmux status bar
-- **PSM (Plugin State Manager)** — SQLite-backed cross-session state persistence with fleet-level visibility
-- **MCP server** — 10 tools for extended capabilities (`omp_get_agents`, `omp_delegate_task`, `omp_activate_skill`, `omp_get_hud_state`, `omp_get_session_state`, `omp_save_session`, `omp_list_sessions`, `omp_invoke_hook`, `omp_subscribe_hud_events`, `omp_fleet_status`)
-- **SWE-bench harness** — reproducible benchmark runner for performance evaluation
-- **Double-tiered MCP config** — user-level (`~/.omp/`) and workspace-level (`.omp/`) config with merge resolution
-- **Setup wizard** — `/setup` and `/mcp-setup` skills for frictionless onboarding
-- **ADR governance** — `archgate` CLI integration for architecture decision records (`d38b46d`, `9cc3d09`)
-- **OMP rename** — project renamed from `oh-my-claudecode (OMC)` to `oh-my-githubcopilot (OMP)` targeting GitHub Copilot CLI (`6ee243f`)
-
-### Documentation
-
-- `AGENTS.md` — agent registry and delegation rules
-- `CLAUDE.md` — project instructions and quick reference
-- `SECURITY.md` — vulnerability reporting policy
-- `FUNDING.yml` — GitHub Sponsors link
-- Spec documents: `spec/AGENTS_SPEC.md`, `spec/SKILLS.md`, `spec/HOOKS.md`, `spec/HUD.md`, `spec/PSM.md`, `spec/MCP.md`
-- 11 localized READMEs, SVG logo, buddy screenshots
+- **Agent parity (23→19):** Six agents removed, merged, or renamed:
+  - `orchestrator` → top-level instruction role (not delegatable)
+  - `researcher` → merged into `document-specialist`
+  - `reviewer` → merged into `code-reviewer`
+  - `tester` → merged into `test-engineer`
+  - `explorer` → renamed to `explore`
+  - `simplifier` → renamed to `code-simplifier`
+- Run `omp doctor` to detect stale agent references in your config files.
+
+### New Features
+
+#### Native Slash Commands (SDK Extension)
+- Every skill is now available as a native `/skill-name` command in GitHub Copilot CLI
+- Aliases: `/ulw` (ultrawork), `/eco` (ecomode), `/plan` (omp-plan), `/di` (deep-interview)
+- Extension registered via `.github/extensions/oh-my-githubcopilot/extension.mjs`
+
+#### Dynamic HUD (Canvas API)
+- Live Canvas HUD updates via Copilot SDK Canvas API (no tmux required)
+- Fallback chain: Canvas → tmux statusline → file polling
+- `OMP_HUD_POLL_MS` env var to tune polling interval
+
+#### 20 New Skills (59 total, was 39)
+**P1:** verify, cancel, help, code-review, security-review, ultraqa, ultragoal  
+**P2:** deep-dive, external-context, deepsearch, sciomc, remember, writer-memory, deepinit  
+**P3:** self-improve, visual-verdict, ccg, build-fix, design, web-clone
+
+#### `omp doctor`
+- Scans `.github/copilot-instructions.md`, `AGENTS.md`, `.omg/`, `.omp/` for stale agent references
+- Exits 1 when issues found (CI-detectable)
+
+### Fixes
+- **Hooks fail-open:** All 6 lifecycle hooks now emit `{"decision":"allow"}` on any error instead of exiting 1 (which caused tool-call denials). `timeoutSec` raised from 0.2→5.
+- **token-tracker Set serialization:** `warnings_issued` persisted as Array, rehydrated as Set — fixes `TypeError: .has is not a function` crash.
+- **Model tier routing:** `analyst`, `designer`, `code-simplifier`, `code-reviewer` correctly routed to `claude-opus-4.6` (high tier).
+- **`/help` and `/cancel` keyword collision:** These common words no longer shadow native Copilot commands.
+
+### Architecture
+- `src/hooks/runner.mts`: shared fail-open entry point for all hooks
+- `src/extension/registry.mts`: single source of truth for all 59 skills
+- `src/cli/doctor.mts`: agent migration scanner
+- ADR-0002: plugin + extension dual-surface architecture
+- ADR-0003: hooks fail-open design
+- ADR-0004: token-tracker Set→Array serialization
+
+## [1.8.1] — (previous release)
+
+See git history for changes prior to 2.0.0.

package/README.md

@@ -40,9 +40,9 @@ omp hud --watch         # live status
 
 | Feature | Description |
 |---------|-------------|
-| **Specialized Agents** | 23+ agents (analyst, architect, executor, debugger, critic, verifier, test-engineer, writer, and more) |
+| **Specialized Agents** | 19 agents (analyst, architect, executor, debugger, critic, verifier, test-engineer, writer, and more) |
 | **Parallel Team Mode** | tmux-based multi-worker orchestration with shared task state |
-| **Workflow Skills** | 39 skills built in — plan, deep-interview, ralph, autopilot, ultrawork, code-review, and more |
+| **Workflow Skills** | 59 skills built in — plan, deep-interview, ralph, autopilot, ultrawork, code-review, and more |
 | **Persistent Hooks** | Automatic tool tracking, project memory, session management |
 | **Real-time HUD** | Live status overlay showing agents, costs, and progress |
 | **CI/CD Ready** | Verification gates, test integration, release workflows |
@@ -138,3 +138,8 @@ omp is open source under the [Apache-2.0 License](LICENSE).
 ## Sponsors
 
 If omp saves you time, consider [sponsoring the project](https://github.com/sponsors/r3dlex) ❤️
+
+<!-- v3-ai-sdlc-init:start -->
+## AI SDLC v3
+This repo follows the v3 AI-SDLC layout. See `.ai/matrix.json`, `.memory/human-override/`, and `docs/architecture/adr/`. Modules at `r3dlex/skills/ai-sdlc-init/modules/`.
+<!-- v3-ai-sdlc-init:end -->

package/agents/code-reviewer.agent.md

@@ -18,6 +18,11 @@ Perform thorough code reviews with severity ratings, SOLID principle checks, and
 - Severity-rated feedback (critical/major/minor)
 - Logic defect detection and SOLID principle violations
 - Style, performance, and maintainability analysis
+- Style enforcement against project linter rules and pattern consistency
+- Merge gate verdicts (approve / request changes)
+
+## Constraints
+- Blocks on security issues, unhandled errors, and type mismatches — not on preference-based style choices outside linter rules
 
 ## NOT Responsible For
 - Implementing fixes (→ executor)

package/agents/{simplifier.agent.md → code-simplifier.agent.md}

@@ -1,5 +1,5 @@
 ---
-name: simplifier
+name: code-simplifier
 description: >
   Code simplification and refactoring specialist.
   Use when: simplifying code, improving clarity, removing dead code, reducing complexity.
@@ -9,7 +9,7 @@ agents: [explore]
 user-invocable: true
 ---
 
-# Simplifier
+# Code Simplifier
 
 ## Role
 Simplify and refine code for clarity, consistency, and maintainability while preserving exact functionality.

package/agents/debugger.agent.md

@@ -3,7 +3,7 @@ name: debugger
 description: >
   Root-cause analysis and failure diagnosis specialist.
   Use when: "debug this", "find the bug", "diagnose failure", stack trace analysis, build errors.
-model: claude-sonnet-4-6
+model: claude-opus-4-6
 tools: [readFile, runInTerminal, search, codebase, problems]
 agents: [explore, architect]
 user-invocable: true

package/agents/document-specialist.agent.md

@@ -2,7 +2,7 @@
 name: document-specialist
 description: >
   External documentation and reference specialist.
-  Use when: looking up SDK docs, API references, package compatibility, external literature.
+  Use when: looking up SDK docs, API references, package compatibility, comparing libraries, researching APIs, external literature.
 model: claude-sonnet-4-6
 tools: [readFile, editFiles, search, fetch]
 agents: [explore]
@@ -18,6 +18,11 @@ Find and synthesize information from the most trustworthy documentation source a
 - External documentation lookup and synthesis
 - API/framework reference research
 - Package evaluation and version compatibility checks
+- Technology comparisons and options benchmarking with objective tradeoffs
+- Source citation: prefer official docs, flag outdated material, cite URLs
+
+## Constraints
+- READ-ONLY for code. Does not implement — finds, synthesizes, and summarizes with cited sources.
 
 ## NOT Responsible For
 - Internal codebase search (→ explore)

package/agents/{explorer.agent.md → explore.agent.md}

@@ -1,5 +1,5 @@
 ---
-name: explorer
+name: explore
 description: >
   Fast codebase surveyor for targeted search and mapping.
   Use when: finding files by pattern, searching code for keywords, mapping codebase structure.
@@ -9,7 +9,7 @@ agents: []
 user-invocable: true
 ---
 
-# Explorer
+# Explore
 
 ## Role
 Perform fast, targeted codebase surveys: find file patterns, map structure, locate symbols, return concise summaries.

package/agents/security-reviewer.agent.md

@@ -3,7 +3,7 @@ name: security-reviewer
 description: >
   Security vulnerability detection specialist (OWASP Top 10, secrets, unsafe patterns).
   Use when: "security review", "find vulnerabilities", "check for secrets", auth/crypto changes.
-model: claude-sonnet-4-6
+model: claude-opus-4-6
 tools: [readFile, search, codebase]
 agents: [explore]
 user-invocable: true