@lazyingart/agintiflow 0.20.25 → 0.20.39

package/skills/devops-deployment/SKILL.md

@@ -0,0 +1,29 @@
+---
+id: devops-deployment
+label: DevOps Deployment And Runtime
+description: Handle Docker, CI/CD, deployment, local services, ports, logs, environment setup, and runtime debugging.
+triggers:
+  - devops
+  - docker
+  - deploy
+  - deployment
+  - ci cd
+  - service
+  - logs
+  - port
+  - environment
+  - nginx
+tools:
+  - inspect_project
+  - read_file
+  - write_file
+  - apply_patch
+  - run_command
+  - web_search
+---
+# DevOps Deployment And Runtime
+
+Diagnose first with read-only evidence: versions, manifests, ports, logs, process status, and existing scripts. Prefer project-local scripts, Docker, or user-writable caches over host mutation.
+
+Make setup idempotent and reversible. For long-running services, use tmux tools where available. If sudo, cloud credentials, or destructive infra changes are required, stop and provide a precise manual step.
+

package/skills/docs-knowledge/SKILL.md

@@ -0,0 +1,28 @@
+---
+id: docs-knowledge
+label: Documentation And Knowledge Base
+description: Write or maintain README files, API docs, tutorials, changelogs, architecture notes, and user-facing documentation.
+triggers:
+  - docs
+  - documentation
+  - readme
+  - tutorial
+  - guide
+  - changelog
+  - api docs
+  - knowledge base
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - write_file
+  - apply_patch
+  - run_command
+  - send_to_canvas
+---
+# Documentation And Knowledge Base
+
+Do not invent behavior when source files can be inspected. Read manifests, exported APIs, CLI help, tests, and examples before documenting commands or options.
+
+Structure docs for a real reader: quick start, concepts, workflows, examples, troubleshooting, and reference. Verify important commands when safe and mark unverified assumptions explicitly.
+

package/skills/dotnet-csharp/SKILL.md

@@ -0,0 +1,24 @@
+---
+id: dotnet-csharp
+label: .NET C# Development
+description: Build, test, debug, and maintain .NET, C#, F#, ASP.NET, console, library, and test projects.
+triggers:
+  - dotnet
+  - .net
+  - c#
+  - csharp
+  - asp.net
+  - nuget
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+---
+# .NET C# Development
+
+Inspect `.sln`, `.csproj`, `Program.cs`, controllers, appsettings, project references, and test projects before editing.
+
+Use `dotnet restore`, `dotnet build`, and `dotnet test` when available. Keep `bin/`, `obj/`, and generated artifacts out of commits, and stop with evidence on missing SDK, private feed, or signing blockers.
+

package/skills/education-tutorial/SKILL.md

@@ -0,0 +1,26 @@
+---
+id: education-tutorial
+label: Education Tutorials And Exercises
+description: Create lessons, tutorials, courses, examples, quizzes, exercises, and learning materials.
+triggers:
+  - education
+  - teach
+  - tutorial
+  - course
+  - lesson
+  - exercise
+  - quiz
+  - homework
+tools:
+  - read_file
+  - write_file
+  - run_command
+  - web_search
+  - send_to_canvas
+---
+# Education Tutorials And Exercises
+
+Start from learner level and objective. Sequence concepts from concrete examples to explanation to practice.
+
+Create durable lesson files with examples, exercises, answer keys, and verification steps. Test code/math examples when practical.
+

package/skills/go/SKILL.md

@@ -0,0 +1,23 @@
+---
+id: go
+label: Go Development
+description: Build, test, debug, and maintain Go modules, CLIs, servers, packages, and tooling.
+triggers:
+  - go
+  - golang
+  - go.mod
+  - go test
+  - gofmt
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+---
+# Go Development
+
+Inspect `go.mod`, commands under `cmd/`, packages under `internal/` or `pkg/`, and tests before editing.
+
+Use `gofmt` on changed Go files and run focused `go test` commands, broadening to `go test ./...` when practical. Keep module/download noise out of commits and report missing Go/private module blockers precisely.
+

package/skills/ios-swift/SKILL.md

@@ -0,0 +1,26 @@
+---
+id: ios-swift
+label: iOS Swift Development
+description: Build, test, debug, and package iOS, Swift, SwiftUI, Xcode, Swift Package Manager, simulator, and Apple-platform projects.
+triggers:
+  - ios
+  - iphone
+  - ipad
+  - swift
+  - swiftui
+  - xcode
+  - simulator
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+  - send_to_canvas
+---
+# iOS Swift Development
+
+Inspect before editing: `Package.swift`, `.xcodeproj`, `.xcworkspace`, schemes, `Info.plist`, assets, signing settings, `xcodebuild -version`, and available simulators.
+
+Prefer Swift Package Manager or project-local Xcode build commands. Verify with `swift test`, `xcodebuild build/test`, simulator launch, or a clear blocker when Xcode/signing/simulator access is unavailable. Save screenshots under durable workspace paths when a simulator is available.
+

package/skills/java-jvm/SKILL.md

@@ -0,0 +1,25 @@
+---
+id: java-jvm
+label: Java JVM Development
+description: Build, test, debug, and maintain Java, Kotlin JVM, Maven, Gradle, Spring, and JUnit projects.
+triggers:
+  - java
+  - jvm
+  - kotlin
+  - maven
+  - gradle
+  - spring
+  - junit
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+---
+# Java JVM Development
+
+Inspect the build system before editing: `pom.xml`, `build.gradle`, `settings.gradle`, wrapper files, modules, source sets, and tests.
+
+Prefer project-local `./mvnw`, `./gradlew`, or existing Maven/Gradle commands. Run focused compile/test tasks first, keep generated `target/`, `build/`, and `.gradle/` artifacts out of commits, and stop on missing JDK or credential blockers with exact evidence.
+

package/skills/php/SKILL.md

@@ -0,0 +1,24 @@
+---
+id: php
+label: PHP Development
+description: Build, test, debug, and maintain PHP, Composer, Laravel, Symfony, WordPress-style, CLI, and web projects.
+triggers:
+  - php
+  - composer
+  - laravel
+  - symfony
+  - phpunit
+  - pest
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+---
+# PHP Development
+
+Inspect `composer.json`, framework config, routes, controllers, migrations, and tests before editing.
+
+Use `php -l`, Composer scripts, PHPUnit, Pest, or framework test commands when available. Avoid global PHP extension or service changes unless explicitly approved, and keep `vendor/` out of commits.
+

package/skills/presentation-slides/SKILL.md

@@ -0,0 +1,26 @@
+---
+id: presentation-slides
+label: Slides And Presentations
+description: Create slide decks, pitch decks, posters, lecture materials, and presentation outlines.
+triggers:
+  - slides
+  - presentation
+  - powerpoint
+  - pptx
+  - deck
+  - pitch deck
+  - poster
+  - lecture
+tools:
+  - read_file
+  - write_file
+  - run_command
+  - send_to_canvas
+  - web_search
+---
+# Slides And Presentations
+
+Clarify audience, goal, length, and tone from context. Build a durable outline before slide text. Keep slides visually sparse: one message, one figure/table/story per slide when possible.
+
+Use markdown, HTML, LaTeX beamer, PPTX tooling, or available local converters depending on the workspace. Save source plus exported artifacts when possible.
+

package/skills/qa-testing/SKILL.md

@@ -0,0 +1,27 @@
+---
+id: qa-testing
+label: QA Testing And CI Debugging
+description: Reproduce bugs, repair failing tests, design regression checks, debug CI, and improve test coverage.
+triggers:
+  - qa
+  - test
+  - testing
+  - failing test
+  - regression
+  - coverage
+  - ci
+  - github actions
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+---
+# QA Testing And CI Debugging
+
+Start by reproducing the failure with the narrowest command when a real failure exists. Capture the exact error and identify whether the failure is test expectation, product bug, setup issue, or flaky timing.
+
+Patch the root cause, not the symptom. Add or update a regression test when it would have caught the issue. Run the focused test again, then a broader suite only when useful.
+
+When the task is to create a QA/testing project from scratch, build a clean runnable project with meaningful tests. Do not invent staged bugs, misleading tests, or artificial failures unless the user explicitly asks for a bug-reproduction fixture. If an interrupted draft exists, remove stale generated artifacts and contradictory comments or README claims before committing.

package/skills/ruby/SKILL.md

@@ -0,0 +1,24 @@
+---
+id: ruby
+label: Ruby Development
+description: Build, test, debug, and maintain Ruby, Rails, Bundler, gems, Rake, and RSpec/Minitest projects.
+triggers:
+  - ruby
+  - rails
+  - gemfile
+  - bundler
+  - rake
+  - rspec
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+---
+# Ruby Development
+
+Inspect `Gemfile`, `.gemspec`, `Rakefile`, Rails routes/models/controllers, migrations, and tests before editing.
+
+Use `bundle exec rake`, `bundle exec rspec`, or focused Ruby test commands when available. Keep `vendor/bundle`, logs, tmp, and generated artifacts out of commits, and report native gem or credential blockers precisely.
+

package/skills/rust/SKILL.md

@@ -0,0 +1,23 @@
+---
+id: rust
+label: Rust Development
+description: Build, test, debug, and maintain Rust crates, Cargo workspaces, CLIs, services, and libraries.
+triggers:
+  - rust
+  - cargo
+  - crate
+  - borrow checker
+  - clippy
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+---
+# Rust Development
+
+Inspect `Cargo.toml`, workspace members, features, modules, examples, and tests before editing.
+
+Prefer narrow `cargo check -p`, `cargo test -p`, or targeted test commands before broad workspace checks. Run `cargo fmt` when available, avoid unnecessary dependency churn, and explain safety/unsafe or lifetime tradeoffs clearly.
+

package/skills/security-review/SKILL.md

@@ -0,0 +1,28 @@
+---
+id: security-review
+label: Security Review
+description: Review secrets, authentication, authorization, dependency risk, input validation, shell safety, and threat models.
+triggers:
+  - security
+  - secret
+  - secrets
+  - auth
+  - authentication
+  - authorization
+  - vulnerability
+  - threat model
+  - dependency audit
+tools:
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+  - web_search
+---
+# Security Review
+
+Never print secrets. Prefer redacted evidence and paths. Distinguish exploitable risk from style issues, and prioritize high-impact fixes.
+
+Inspect auth boundaries, file/path handling, shell command construction, dependency manifests, and config defaults. Run available audit tools when safe, but do not treat scanners as a substitute for code review.
+

package/skills/supervision-student/SKILL.md

@@ -0,0 +1,67 @@
+---
+id: supervision-student
+label: Supervision Student Loop
+description: Supervise another AgInTiFlow/student agent through tmux or session logs, verify its work independently, and convert failures into durable capability improvements.
+triggers:
+  - supervision
+  - supervise
+  - supervisor
+  - student agent
+  - homework
+  - curriculum
+  - self supervision
+  - train agent
+  - tmux monitor
+tools:
+  - tmux_list_sessions
+  - tmux_capture_pane
+  - tmux_send_keys
+  - tmux_start_session
+  - inspect_project
+  - search_files
+  - read_file
+  - apply_patch
+  - run_command
+  - send_to_canvas
+---
+# Supervision Student Loop
+
+Use this skill when AgInTiFlow is supervising another agent, a tmux session, or a long-running autonomous task.
+
+## Roles
+
+- The student does the target project work.
+- The supervisor gives normal user-level prompts, monitors progress, verifies evidence, and improves reusable capabilities.
+- The supervisor should not directly implement the student's target project unless the user explicitly changes the assignment.
+
+## Loop
+
+1. Create or reuse a clean project folder and a named tmux/session id.
+2. Give the student a realistic task with imperfect prompting. Do not over-specify every implementation step.
+3. Define acceptance criteria before judging: files, build/test output, screenshots, commits, docs, UX, or source evidence.
+4. Monitor progress through tmux capture, session events, runtime logs, git status, and artifact lists.
+5. Verify the result externally. Do not trust the student's final summary by itself.
+6. If the student is blocked by missing reusable capability, improve AgInTiFlow itself: profile, skill, tool, command policy, model prompt, UI, docs, or smoke test.
+7. Resume the same student session with the updated AgInTiFlow and ask it to continue.
+8. Record the outcome in a homework ledger with evidence, flaws, fixes, and next test.
+
+## Evidence Checklist
+
+- `git status --short` before and after.
+- `git diff --stat` or commit log for intended changes.
+- Build/test/check output from the actual toolchain.
+- Durable output paths for screenshots, images, PDFs, reports, APKs, archives, and generated docs.
+- Session `events.jsonl` snippets when tool use or artifact publication matters.
+- A short postmortem: what failed, what capability was added, and what to test next.
+
+## Recursive Supervision
+
+When training a supervisor agent, give it a smaller student to supervise. Grade the supervisor on whether it:
+
+- Sets acceptance criteria.
+- Gives compact prompts instead of doing the work.
+- Verifies artifacts independently.
+- Notices loops, false claims, missing files, bad git state, and tool failures.
+- Converts recurring failures into reusable improvements.
+
+The goal is not only a better worker. The goal is an agent that can supervise work, discover its own capability gaps, and request or implement durable upgrades.

package/skills/writing-editing/SKILL.md

@@ -0,0 +1,28 @@
+---
+id: writing-editing
+label: Writing Editing And Scriptwriting
+description: Draft, edit, rewrite, structure, and polish articles, markdown, scripts, screenplays, books, novels, and creative or professional prose.
+triggers:
+  - writing
+  - edit
+  - rewrite
+  - markdown
+  - blog
+  - article
+  - copywriting
+  - screenplay
+  - final draft
+  - script writing
+tools:
+  - read_file
+  - write_file
+  - apply_patch
+  - web_search
+  - send_to_canvas
+---
+# Writing Editing And Scriptwriting
+
+Preserve the user's voice and purpose. For long work, create or update durable files: outline, draft, revision notes, style guide, and continuity notes.
+
+For editing, identify the target audience, tone, format, and acceptance criteria before changing structure. For scripts/screenplays, track scenes, beats, characters, dialogue intent, and revision history.
+

package/src/agent-runner.js

@@ -15,7 +15,7 @@ import { normalizeWrapperName, runAgentWrapper, wrapperStatusText } from "./tool
 import { evaluateCommandPolicy } from "./command-policy.js";
 import { redactSensitiveText, redactValue } from "./redaction.js";
 import { executeWorkspaceTool, resolveWorkspacePath, summarizeWorkspaceTools, WORKSPACE_TOOL_NAMES } from "./workspace-tools.js";
-import { normalizeCanvasPayload } from "./artifact-tunnel.js";
+import { normalizeCanvasPayload, persistCanvasPayloadFile } from "./artifact-tunnel.js";
 import { getTaskProfile } from "./task-profiles.js";
 import { generateImage, listAuxiliarySkills } from "./auxiliary-tools.js";
 import { engineeringGuidanceForTask } from "./engineering-guidance.js";
@@ -268,6 +268,8 @@ async function createInitialState(config, sessionId) {
     provider: config.provider,
     model: config.model,
     goal: config.goal,
+    baseDir: config.baseDir,
+    commandCwd: config.commandCwd,
     startUrl: config.startUrl,
     plan: "",
     stepsCompleted: 0,
@@ -332,14 +334,14 @@ async function createInitialState(config, sessionId) {
           `Task profile: ${taskProfile.label}. ${taskProfile.prompt}`,
           skillContext,
           engineeringGuidance,
-          "A frontend canvas/artifacts tunnel exists. Use send_to_canvas when important markdown, diffs, screenshots, images, or workspace files should be highlighted in the UI. It is optional and ordinary final text can still go directly to finish.",
+          "A frontend canvas/artifacts tunnel exists. Use send_to_canvas when important markdown, diffs, screenshots, images, or workspace files should be highlighted in the UI. File paths sent to canvas are copied into session artifacts for durable preview, but user-requested outputs should also remain in a clear workspace path unless the user asked only for a temporary preview.",
           "For visual-output requests such as draw, plot, graph, chart, diagram, figure, image, or visualization, proactively publish a canvas artifact even when the user does not mention canvas. If workspace file tools are enabled, prefer creating a small SVG or markdown artifact and call send_to_canvas with selected=true.",
           "Work like a practical coding agent: orient with inspect_project/search/read, patch code with apply_patch, run safe checks when they add confidence, iterate on failures, and keep outputs inside the workspace.",
           "For large projects, decompose into useful files and milestones, identify entry points/tests/contracts first, implement a coherent minimal version, then iterate with checks rather than only describing what you would do.",
           "For website/app/code/LaTeX/Python/C/shell tasks, create or edit real workspace files, run available build/compile/test commands, and surface artifacts through the canvas when useful.",
           "For LaTeX/PDF tasks, check existing latexmk/pdflatex first and compile with the available host or Docker TeX toolchain before installing packages or rebuilding the sandbox.",
           "For research or web-search tasks, use browser tools or safe shell network tools when the current policy allows; cite or save useful sources in workspace notes when the task needs traceability.",
-          "Use the canvas tunnel for outputs the user would likely want to inspect visually, such as figures, PDFs, screenshots, images, important markdown, or generated files.",
+          "Use the canvas tunnel for outputs the user would likely want to inspect visually, such as figures, PDFs, screenshots, images, important markdown, or generated files. When no save path is specified, choose a descriptive non-conflicting workspace path near the working directory and keep it there.",
           "For environment or system-maintenance work, use the configured sandbox and package policy; Docker workspace mode is the preferred place for installs and toolchain setup.",
           "For long-running work, create a durable checkpoint or artifact at each completed phase, then continue with the next concrete phase until the requested outcome is actually complete or blocked by a real dependency.",
           "If the user asks to open a generated local website or file, use open_workspace_file for a file or preview_workspace for a static site. Do not keep retrying the same localhost URL when a preview fails.",
@@ -737,7 +739,7 @@ async function captureSyntheticSnapshot(store, step, config) {
       config.allowWrapperTools
         ? `Agent wrappers available: selected=${normalizeWrapperName(config.preferredWrapper)}; ${wrapperStatusText()}`
         : "Agent wrappers disabled.",
-      "Canvas/artifacts tunnel available through send_to_canvas.",
+      "Canvas/artifacts tunnel available through send_to_canvas. File paths sent to canvas are persisted into the session artifact store, but final user artifacts should still use clear durable workspace filenames.",
       "For draw/plot/graph/chart/diagram/figure requests, publish a canvas artifact proactively.",
       "For LaTeX/PDF requests, check latexmk/pdflatex first, publish the source and compiled PDF artifacts when available, and avoid reinstalling TeX when an existing toolchain works.",
       "Use open_url only if the task actually needs the web.",
@@ -1097,8 +1099,22 @@ async function executeTool(browserState, toolCall, snapshot, config, store, obse
               config
             );
             if (normalized.ok) {
+              const persisted = await persistCanvasPayloadFile(normalized.payload, { config, store });
+              if (!persisted.ok) {
+                await store.appendEvent("canvas.persistence_failed", {
+                  toolName: "generate_image",
+                  reason: persisted.reason,
+                  path: normalized.payload.path,
+                });
+                observers.event("canvas.persistence_failed", {
+                  toolName: "generate_image",
+                  reason: persisted.reason,
+                  path: normalized.payload.path,
+                });
+                return result;
+              }
               const canvasItem = {
-                ...normalized.payload,
+                ...persisted.payload,
                 toolName: "generate_image",
                 commandCwd: config.commandCwd,
               };
@@ -1144,8 +1160,30 @@ async function executeTool(browserState, toolCall, snapshot, config, store, obse
           };
         }
 
+        const persisted = await persistCanvasPayloadFile(normalized.payload, { config, store });
+        if (!persisted.ok) {
+          await store.appendEvent("tool.blocked", {
+            toolName: "send_to_canvas",
+            args: safeArgs,
+            reason: persisted.reason,
+            category: "canvas",
+          });
+          observers.event("tool.blocked", {
+            toolName: "send_to_canvas",
+            args: safeArgs,
+            reason: persisted.reason,
+            category: "canvas",
+          });
+          return {
+            ok: false,
+            blocked: true,
+            reason: persisted.reason,
+            toolName: "send_to_canvas",
+          };
+        }
+
         const canvasItem = {
-          ...normalized.payload,
+          ...persisted.payload,
           toolName: "send_to_canvas",
           commandCwd: config.commandCwd,
         };
@@ -1216,7 +1254,10 @@ export async function runAgent(config) {
   }
 
   const sessionId = config.resume || config.sessionId || `web-agent-${crypto.randomUUID()}`;
-  const store = new SessionStore(config.sessionsDir, sessionId);
+  const store = new SessionStore(config.sessionsDir, sessionId, {
+    projectRoot: config.baseDir,
+    projectSessionsDir: config.projectSessionsDir,
+  });
   const client = createClient(config);
   const observers = createObservers(config);
   const browserState = createBrowserState();
@@ -1353,6 +1394,7 @@ export async function runAgent(config) {
     emitConsole(config, `Routing: ${config.routingMode} (${config.routeReason})`, { kind: "meta" });
     emitConsole(config, `Workspace: ${config.commandCwd}`, { kind: "meta" });
     emitConsole(config, `Sessions: ${config.sessionsDir}`, { kind: "meta" });
+    if (config.projectSessionsDir) emitConsole(config, `Project session index: ${config.projectSessionsDir}`, { kind: "meta" });
     if (config.useDockerSandbox) {
       emitConsole(
         config,