v0.13.4

Release Notes (v0.13.3 → v0.13.4)

🎉 New Features

Configurable Effort Levels (#959)

• New --effort flag (low / medium / high / max) on il start, il spin, and il plan controls how deeply agents reason through each phase
• Resolution chain lets you set effort at multiple levels: CLI flag > loom metadata > settings > environment variable > default
• Swarm mode uses sensible per-agent defaults (analyzer=high, implementer=medium) and supports separate swarmEffort overrides in settings

Dockerfile Detection in Config Wizard (#964)

• il init now detects Dockerfiles in your project and offers to configure Docker-based dev servers automatically
• Saves a manual configuration step for containerized projects

Code Review Fix Tracking (#961)

• New fix recap entry type lets agents log when a code review finding has been addressed
• Creates a clear audit trail from review finding → resolution in the loom recap

🐛 Bug Fixes

OAuth Authentication Broken for Branch/Commit Generation (#965)

• An internal environment variable was disabling OAuth credential lookup in Claude CLI, causing "Not logged in" errors when generating branch names or commit messages without an explicit API key

---

Full Changelog: v0.13.3...v0.13.4

v0.13.3

Release Notes (v0.12.0 → v0.13.3)

🎉 New Features

Composable Automation Flags for il plan (#908, #909)

• --yolo has been decomposed into independent, composable flags so you can fine-tune how much autonomy the planner gets
• --one-shot=noReview skips confirmation gates but keeps permission prompts; --one-shot=bypassPermissions skips both
• --dangerously-skip-permissions is now available standalone on both il plan and il start
• --auto-swarm is independent — it no longer implies full autonomy
• --yolo remains as a convenient superset shorthand

Epic Swarm Implementation Report (#928, #934)

• When finishing an epic loom, iloom now generates a comprehensive implementation report covering what each child issue delivered, integration points to review carefully, and lessons learned
• Reports are posted as PR comments with collapsible sections so reviewers can skim the summary and drill into per-issue details
• Includes testing guidance and a verification checklist derived from actual implementation data — not boilerplate

Dev Server TUI with Sticky Status Bar (#922, #923, #955)

• il dev-server now shows a persistent status bar at the bottom of the terminal with the external URL, container status, and port mapping
• Keyboard shortcuts: o (open browser), c (copy URL), r (restart), q (quit)
• Zero new dependencies — activates automatically in TTY environments, falls through to existing behavior otherwise

1M Context Window Model Support (#907)

• opus[1m] and sonnet[1m] model selections are now available across all commands
• Swarm orchestrator defaults to opus[1m] in both Balanced and Maximum Quality modes for deeper reasoning over large codebases

Docker BuildKit Secrets and Runtime Environment Variables (#919, #935)

• New dockerBuildSecrets config maps secret IDs to local files, passed as --secret flags during docker build
• New dockerRunEnv setting and repeatable -e, --env KEY=VALUE CLI flag on dev-server, open, and run commands for passing environment variables to Docker containers at runtime
• Clear precedence: settings < .env files < CLI flags (last wins)
• Graceful container shutdown on Ctrl+C no longer produces noisy error output

Configurable Dev Server Protocol (#921)

• New protocol option (http/https) in web capabilities settings so dev server URLs display the correct scheme across il open, il run, and il dev-server

Smarter Planning with Dedicated Research Agent (#879, #880)

• Plan mode now delegates research to the Analyze agent instead of doing inline discovery, producing deeper and more thorough implementation plans
• Plans benefit from the same structured research methodology used during issue implementation

Integration-Point Must-Haves in Plans (#912)

• The planner now detects producer/consumer relationships between parallel child issues in the same wave and generates substantive must-haves on verification issues
• Ensures the verifier checks that consumers actually use real instances instead of null placeholders after parallel work merges

Pre-Trusted Tools in Plan Mode (#869, #900)

• Safe, read-only tools and issue management tools are pre-trusted during planning, reducing approval prompts so you can focus on reviewing the plan

Wave Verifier Enhancements (#899, #902, #911)

• Verification results are now posted as comments directly on the verifier's issue, creating a visible audit trail
• Code review step added to the wave verification workflow — gathers the wave diff, invokes the reviewer, and auto-fixes critical findings
• Fix step now parses findings by file and invokes one implementer per file in parallel

Swarm Visibility Improvements (#901, #903, #950)

• Full agent routing table displayed after swarm completes so you can see how every issue was handled
• Swarm comment strategy switched from incremental updates to a single final summary, reducing overhead
• Epic dev server port forwarded to the orchestrator for web output verification after merging child work

Structured Draft PR Bodies (#953, #954)

• Draft PRs created by il start now include the issue title as a heading and the issue body in a collapsible details section, giving reviewers immediate context

Streamlined VS Code Init Wizard (#952)

• When running from VS Code, the config wizard detects the IDE automatically and skips the IDE selection question

Auto-Ignore .env.local Files (#927)

• Local environment override files (.env.local, .env.development.local, etc.) are now automatically added to the global gitignore on upgrade, preventing accidental commits of secrets across all your projects

🐛 Bug Fixes

il list Now Shows Capabilities for Main Worktree (#958)

• il list --json returned empty capabilities for the main worktree because it only reads from loom metadata — now detects capabilities at runtime using the same logic as il dev

Wave Verification Recap Entries Now Saved Correctly (#948, #949)

• Code reviewer agents in swarm mode weren't receiving worktree paths, causing recap entries to route to the wrong file or not be created — now fixed with proper path context

Dev Server Respects Custom Commands in TUI Mode (#926)

• il dev now uses the dev command from package.iloom.local.json in all modes — previously the TUI code path ignored it and fell back to pnpm dev

Docker Container Name Conflicts When Running Multiple Looms (#926)

• Running il dev for multiple Docker-based looms concurrently no longer fails with "container name already in use" — stale containers are now verified as removed before starting new ones

Agent Loading Resilient to Non-Agent Files (#956)

• Documentation files in the agents directory no longer prevent all agents from loading — they're skipped gracefully

Rebase Conflict Resolution Runs Smoothly (#925)

• Read-only tools are now pre-granted during rebase conflict resolution, so the assistant doesn't pause for approval on routine file reads

Main Worktree Uses Base Port (#924)

• Running il dev-server, il open, or il run on the main worktree now correctly returns the base port (default 3000) instead of a hash-derived offset

Plan Agent Respects Configured Issue Tracker (#904)

• The planner now knows which issue tracker and VCS provider are configured, preventing it from using gh CLI commands when the tracker is Linear or Jira

Clarifying Questions Batched (#914)

• The planner now batches up to 4 independent clarifying questions per round instead of asking one at a time

il list Settings Resolution (#910)

• Running il list from a child worktree with a settings.local.json overriding mainBranch no longer filters out all looms

Branch Mode Reuses Existing Branches (#913)

• il start in branch mode no longer fails when the named branch already exists locally — it reuses the existing branch

Cleanup Supports Alphanumeric Identifiers (#898)

• The --issue flag on il cleanup now handles Linear/Jira-style IDs like RANDOM-13 instead of silently failing

Swarm Child Worktree Matching (#897)

• Cleanup now finds swarm child worktrees using both issue-{id} and issue/{id} branch patterns

VS Code "Create Loom" with Short Titles (#884)

• Creating a loom from VS Code with a short title and a description body no longer fails with "Invalid branch name"

---

Full Changelog: v0.12.0...v0.13.3

v0.13.2

Release Notes (v0.12.0 → v0.13.2)

🎉 New Features

Composable Automation Flags for il plan (#908, #909)

• --yolo has been decomposed into independent, composable flags so you can fine-tune how much autonomy the planner gets
• --one-shot=noReview skips confirmation gates but keeps permission prompts; --one-shot=bypassPermissions skips both
• --dangerously-skip-permissions is now available standalone on both il plan and il start
• --auto-swarm is independent — it no longer implies full autonomy
• --yolo remains as a convenient superset shorthand

Epic Swarm Implementation Report (#928, #934)

• When finishing an epic loom, iloom now generates a comprehensive implementation report covering what each child issue delivered, integration points to review carefully, and lessons learned
• Reports are posted as PR comments with collapsible sections so reviewers can skim the summary and drill into per-issue details
• Includes testing guidance and a verification checklist derived from actual implementation data — not boilerplate

Dev Server TUI with Sticky Status Bar (#922, #923, #955)

• il dev-server now shows a persistent status bar at the bottom of the terminal with the external URL, container status, and port mapping
• Keyboard shortcuts: o (open browser), c (copy URL), r (restart), q (quit)
• Zero new dependencies — activates automatically in TTY environments, falls through to existing behavior otherwise

1M Context Window Model Support (#907)

• opus[1m] and sonnet[1m] model selections are now available across all commands
• Swarm orchestrator defaults to opus[1m] in both Balanced and Maximum Quality modes for deeper reasoning over large codebases

Docker BuildKit Secrets and Runtime Environment Variables (#919, #935)

• New dockerBuildSecrets config maps secret IDs to local files, passed as --secret flags during docker build
• New dockerRunEnv setting and repeatable -e, --env KEY=VALUE CLI flag on dev-server, open, and run commands for passing environment variables to Docker containers at runtime
• Clear precedence: settings < .env files < CLI flags (last wins)
• Graceful container shutdown on Ctrl+C no longer produces noisy error output

Configurable Dev Server Protocol (#921)

• New protocol option (http/https) in web capabilities settings so dev server URLs display the correct scheme across il open, il run, and il dev-server

Smarter Planning with Dedicated Research Agent (#879, #880)

• Plan mode now delegates research to the Analyze agent instead of doing inline discovery, producing deeper and more thorough implementation plans
• Plans benefit from the same structured research methodology used during issue implementation

Integration-Point Must-Haves in Plans (#912)

• The planner now detects producer/consumer relationships between parallel child issues in the same wave and generates substantive must-haves on verification issues
• Ensures the verifier checks that consumers actually use real instances instead of null placeholders after parallel work merges

Pre-Trusted Tools in Plan Mode (#869, #900)

• Safe, read-only tools and issue management tools are pre-trusted during planning, reducing approval prompts so you can focus on reviewing the plan

Wave Verifier Enhancements (#899, #902, #911)

• Verification results are now posted as comments directly on the verifier's issue, creating a visible audit trail
• Code review step added to the wave verification workflow — gathers the wave diff, invokes the reviewer, and auto-fixes critical findings
• Fix step now parses findings by file and invokes one implementer per file in parallel

Swarm Visibility Improvements (#901, #903, #950)

• Full agent routing table displayed after swarm completes so you can see how every issue was handled
• Swarm comment strategy switched from incremental updates to a single final summary, reducing overhead
• Epic dev server port forwarded to the orchestrator for web output verification after merging child work

Structured Draft PR Bodies (#953, #954)

• Draft PRs created by il start now include the issue title as a heading and the issue body in a collapsible details section, giving reviewers immediate context

Streamlined VS Code Init Wizard (#952)

• When running from VS Code, the config wizard detects the IDE automatically and skips the IDE selection question

Auto-Ignore .env.local Files (#927)

• Local environment override files (.env.local, .env.development.local, etc.) are now automatically added to the global gitignore on upgrade, preventing accidental commits of secrets across all your projects

🐛 Bug Fixes

Wave Verification Recap Entries Now Saved Correctly (#948, #949)

• Code reviewer agents in swarm mode weren't receiving worktree paths, causing recap entries to route to the wrong file or not be created — now fixed with proper path context

Dev Server Respects Custom Commands in TUI Mode (#926)

• il dev now uses the dev command from package.iloom.local.json in all modes — previously the TUI code path ignored it and fell back to pnpm dev

Docker Container Name Conflicts When Running Multiple Looms (#926)

• Running il dev for multiple Docker-based looms concurrently no longer fails with "container name already in use" — stale containers are now verified as removed before starting new ones

Agent Loading Resilient to Non-Agent Files (#956)

• Documentation files in the agents directory no longer prevent all agents from loading — they're skipped gracefully

Rebase Conflict Resolution Runs Smoothly (#925)

• Read-only tools are now pre-granted during rebase conflict resolution, so the assistant doesn't pause for approval on routine file reads

Main Worktree Uses Base Port (#924)

• Running il dev-server, il open, or il run on the main worktree now correctly returns the base port (default 3000) instead of a hash-derived offset

Plan Agent Respects Configured Issue Tracker (#904)

• The planner now knows which issue tracker and VCS provider are configured, preventing it from using gh CLI commands when the tracker is Linear or Jira

Clarifying Questions Batched (#914)

• The planner now batches up to 4 independent clarifying questions per round instead of asking one at a time

il list Settings Resolution (#910)

• Running il list from a child worktree with a settings.local.json overriding mainBranch no longer filters out all looms

Branch Mode Reuses Existing Branches (#913)

• il start in branch mode no longer fails when the named branch already exists locally — it reuses the existing branch

Cleanup Supports Alphanumeric Identifiers (#898)

• The --issue flag on il cleanup now handles Linear/Jira-style IDs like RANDOM-13 instead of silently failing

Swarm Child Worktree Matching (#897)

• Cleanup now finds swarm child worktrees using both issue-{id} and issue/{id} branch patterns

VS Code "Create Loom" with Short Titles (#884)

• Creating a loom from VS Code with a short title and a description body no longer fails with "Invalid branch name"

---

Full Changelog: v0.12.0...v0.13.2

v0.13.1

Release Notes (v0.12.0 → v0.13.1)

🎉 New Features

Composable Automation Flags for il plan (#908, #909)

• --yolo has been decomposed into independent, composable flags so you can fine-tune how much autonomy the planner gets
• --one-shot=noReview skips confirmation gates but keeps permission prompts; --one-shot=bypassPermissions skips both
• --dangerously-skip-permissions is now available standalone on both il plan and il start
• --auto-swarm is independent — it no longer implies full autonomy
• --yolo remains as a convenient superset shorthand

Dev Server TUI with Sticky Status Bar (#922, #923)

• il dev-server now shows a persistent status bar at the bottom of the terminal with the external URL, container status, and port mapping
• Keyboard shortcuts: o (open browser), c (copy URL), r (restart), q (quit)
• Zero new dependencies — activates automatically in TTY environments, falls through to existing behavior otherwise

1M Context Window Model Support (#907)

• opus[1m] and sonnet[1m] model selections are now available across all commands
• Swarm orchestrator defaults to opus[1m] in both Balanced and Maximum Quality modes for deeper reasoning over large codebases

Docker BuildKit Secrets (#919)

• New dockerBuildSecrets config maps secret IDs to local files, passed as --secret flags during docker build
• Secrets are validated (file existence, no directory traversal) and DOCKER_BUILDKIT=1 is set automatically
• Graceful container shutdown on Ctrl+C no longer produces noisy error output

Configurable Dev Server Protocol (#921)

• New protocol option (http/https) in web capabilities settings so dev server URLs display the correct scheme across il open, il run, and il dev-server

Smarter Planning with Dedicated Research Agent (#879, #880)

• Plan mode now delegates research to the Analyze agent instead of doing inline discovery, producing deeper and more thorough implementation plans
• Plans benefit from the same structured research methodology used during issue implementation

Integration-Point Must-Haves in Plans (#912)

• The planner now detects producer/consumer relationships between parallel child issues in the same wave and generates substantive must-haves on verification issues
• Ensures the verifier checks that consumers actually use real instances instead of null placeholders after parallel work merges

Pre-Trusted Tools in Plan Mode (#869, #900)

• Safe, read-only tools and issue management tools are pre-trusted during planning, reducing approval prompts so you can focus on reviewing the plan

Wave Verifier Enhancements (#899, #902, #911)

• Verification results are now posted as comments directly on the verifier's issue, creating a visible audit trail
• Code review step added to the wave verification workflow — gathers the wave diff, invokes the reviewer, and auto-fixes critical findings
• Fix step now parses findings by file and invokes one implementer per file in parallel

Swarm Visibility Improvements (#901, #903)

• Full agent routing table displayed after swarm completes so you can see how every issue was handled
• Swarm comment strategy switched from incremental updates to a single final summary, reducing overhead

Auto-Ignore .env.local Files (#927)

• Local environment override files (.env.local, .env.development.local, etc.) are now automatically added to the global gitignore on upgrade, preventing accidental commits of secrets across all your projects

🐛 Bug Fixes

Dev Server Respects Custom Commands in TUI Mode (#926)

• il dev now uses the dev command from package.iloom.local.json in all modes — previously the TUI code path ignored it and fell back to pnpm dev

Docker Container Name Conflicts When Running Multiple Looms (#926)

• Running il dev for multiple Docker-based looms concurrently no longer fails with "container name already in use" — stale containers are now verified as removed before starting new ones

Dev Server TUI Clears Previous Output (#926)

• The TUI no longer overwrites existing terminal content (build logs, startup messages) — it clears the screen first for a clean display

Rebase Conflict Resolution Runs Smoothly (#925)

• Read-only tools are now pre-granted during rebase conflict resolution, so the assistant doesn't pause for approval on routine file reads

Main Worktree Uses Base Port (#924)

• Running il dev-server, il open, or il run on the main worktree now correctly returns the base port (default 3000) instead of a hash-derived offset

Plan Agent Respects Configured Issue Tracker (#904)

• The planner now knows which issue tracker and VCS provider are configured, preventing it from using gh CLI commands when the tracker is Linear or Jira

Clarifying Questions Batched (#914)

• The planner now batches up to 4 independent clarifying questions per round instead of asking one at a time

il list Settings Resolution (#910)

• Running il list from a child worktree with a settings.local.json overriding mainBranch no longer filters out all looms

Branch Mode Reuses Existing Branches (#913)

• il start in branch mode no longer fails when the named branch already exists locally — it reuses the existing branch

Cleanup Supports Alphanumeric Identifiers (#898)

• The --issue flag on il cleanup now handles Linear/Jira-style IDs like RANDOM-13 instead of silently failing

Swarm Child Worktree Matching (#897)

• Cleanup now finds swarm child worktrees using both issue-{id} and issue/{id} branch patterns

VS Code "Create Loom" with Short Titles (#884)

• Creating a loom from VS Code with a short title and a description body no longer fails with "Invalid branch name"

---

Full Changelog: v0.12.0...v0.13.1

v0.13.0

Release Notes (v0.12.0 → v0.13.0)

🎉 New Features

Composable Automation Flags for il plan (#908, #909)

• --yolo has been decomposed into independent, composable flags so you can fine-tune how much autonomy the planner gets
• --one-shot=noReview skips confirmation gates but keeps permission prompts; --one-shot=bypassPermissions skips both
• --dangerously-skip-permissions is now available standalone on both il plan and il start
• --auto-swarm is independent — it no longer implies full autonomy
• --yolo remains as a convenient superset shorthand

Dev Server TUI with Sticky Status Bar (#922, #923)

• il dev-server now shows a persistent status bar at the bottom of the terminal with the external URL, container status, and port mapping
• Keyboard shortcuts: o (open browser), c (copy URL), r (restart), q (quit)
• Zero new dependencies — activates automatically in TTY environments, falls through to existing behavior otherwise

1M Context Window Model Support (#907)

• opus[1m] and sonnet[1m] model selections are now available across all commands
• Swarm orchestrator defaults to opus[1m] in both Balanced and Maximum Quality modes for deeper reasoning over large codebases

Docker BuildKit Secrets (#919)

• New dockerBuildSecrets config maps secret IDs to local files, passed as --secret flags during docker build
• Secrets are validated (file existence, no directory traversal) and DOCKER_BUILDKIT=1 is set automatically
• Graceful container shutdown on Ctrl+C no longer produces noisy error output

Configurable Dev Server Protocol (#921)

• New protocol option (http/https) in web capabilities settings so dev server URLs display the correct scheme across il open, il run, and il dev-server

Smarter Planning with Dedicated Research Agent (#879, #880)

• Plan mode now delegates research to the Analyze agent instead of doing inline discovery, producing deeper and more thorough implementation plans
• Plans benefit from the same structured research methodology used during issue implementation

Integration-Point Must-Haves in Plans (#912)

• The planner now detects producer/consumer relationships between parallel child issues in the same wave and generates substantive must-haves on verification issues
• Ensures the verifier checks that consumers actually use real instances instead of null placeholders after parallel work merges

Pre-Trusted Tools in Plan Mode (#869, #900)

• Safe, read-only tools and issue management tools are pre-trusted during planning, reducing approval prompts so you can focus on reviewing the plan

Wave Verifier Enhancements (#899, #902, #911)

• Verification results are now posted as comments directly on the verifier's issue, creating a visible audit trail
• Code review step added to the wave verification workflow — gathers the wave diff, invokes the reviewer, and auto-fixes critical findings
• Fix step now parses findings by file and invokes one implementer per file in parallel

Swarm Visibility Improvements (#901, #903)

• Full agent routing table displayed after swarm completes so you can see how every issue was handled
• Swarm comment strategy switched from incremental updates to a single final summary, reducing overhead

🐛 Bug Fixes

Main Worktree Uses Base Port (#924)

• Running il dev-server, il open, or il run on the main worktree now correctly returns the base port (default 3000) instead of a hash-derived offset

Plan Agent Respects Configured Issue Tracker (#904)

• The planner now knows which issue tracker and VCS provider are configured, preventing it from using gh CLI commands when the tracker is Linear or Jira

Clarifying Questions Batched (#914)

• The planner now batches up to 4 independent clarifying questions per round instead of asking one at a time

il list Settings Resolution (#910)

• Running il list from a child worktree with a settings.local.json overriding mainBranch no longer filters out all looms

Branch Mode Reuses Existing Branches (#913)

• il start in branch mode no longer fails when the named branch already exists locally — it reuses the existing branch

Cleanup Supports Alphanumeric Identifiers (#898)

• The --issue flag on il cleanup now handles Linear/Jira-style IDs like RANDOM-13 instead of silently failing

Swarm Child Worktree Matching (#897)

• Cleanup now finds swarm child worktrees using both issue-{id} and issue/{id} branch patterns

VS Code "Create Loom" with Short Titles (#884)

• Creating a loom from VS Code with a short title and a description body no longer fails with "Invalid branch name"

---

Full Changelog: v0.12.0...v0.13.0

v0.12.1

Release Notes (v0.12.0 → v0.12.1)

🎉 New Features

Smarter Planning with Dedicated Research Agent (#879, #880)

• Plan mode now delegates research to the Analyze agent instead of doing inline discovery, producing deeper and more thorough implementation plans
• Plans benefit from the same structured research methodology used during issue implementation — covering problem space, third-party libraries, and codebase analysis

Pre-Trusted Tools in Plan Mode (#869)

• A curated set of safe, read-only tools are now pre-trusted during planning, reducing approval prompts so you can focus on reviewing the plan itself

🐛 Bug Fixes

VS Code "Create Loom" Fails for Short Titles (#884)

• Creating a new loom from the VS Code extension with a short issue title (e.g., "A Test Loom") and a description body no longer fails with "Invalid branch name"
• The CLI now correctly recognizes short titles as new issue descriptions when a --body is provided

---

Full Changelog: v0.12.0...v0.12.1

v0.12.0

Release Notes (v0.11.0 → v0.12.0)

🎉 New Features

Wave Verification for Swarm Mode (#836)

• Swarm orchestration now validates work at dependency wave boundaries before spawning the next wave
• The planner generates must-have criteria (exists, substantive, wired) for each child issue, and a verifier agent checks them after each wave completes
• Failed checks trigger automatic fix agents with single-pass re-verification — catching integration issues early instead of at the end
• Verification tasks appear as child issues in the DAG, giving you full visibility into what was checked and what was fixed
• Controlled via waveVerification plan setting (enabled by default)

BitBucket Cloud Integration (#609)

• iloom now supports BitBucket Cloud as a version control provider alongside GitHub
• Create and manage pull requests, post inline review comments, and auto-assign reviewers — all through the same il finish workflow
• Auto-detects workspace and repository from your git remote
• Existing merge mode settings (github-pr, github-draft-pr) are automatically migrated to generic values (pr, draft-pr) so your config works with any provider
• Thanks to @NoahCardoza for the contribution

Docker-Based Dev Server Support (#754)

• Run your dev server inside a Docker container with automatic port mapping via devServer: "docker"
• 3-tier container port resolution: iloom config → Docker image inspect → Dockerfile EXPOSE
• Automatic image building, container lifecycle management, and signal forwarding
• Early crash detection so you know immediately when the container fails to start

Swarm Agents Run as Skills Instead of Subprocesses (#863)

• Swarm phase agents (analyzer, planner, implementer, etc.) now run as native Claude Code skills instead of claude -p subprocesses
• More reliable execution with proper context passing and error visibility

🐛 Bug Fixes

Terminal Corruption on Ctrl+C (#882)

• Fixed terminal corruption when pressing Ctrl+C during foreground child processes — cleanup now runs properly before the process exits

set_goal Firing in Wrong Workflows (#885)

• Goal-setting is now correctly restricted to PR workflows only, preventing unintended behavior in issue and epic looms

Swarm Reliability Improvements

• Increased agent turn limits to prevent premature cutoffs on complex tasks (#860)
• Default swarm worker model corrected to Sonnet (was incorrectly falling back to Opus)

---

Full Changelog: v0.11.0...v0.12.0

v0.11.1

Release Notes (v0.10.0 → v0.11.1)

🎉 New Features

Linux, WSL, and tmux Terminal Support

• iloom now works on Linux with GUI terminals (gnome-terminal, konsole, xterm), WSL via Windows Terminal, and headless environments via tmux fallback (#796)
• Auto-detection: tries GUI terminals first, falls back to tmux when no display is available
• Unblocks il start on headless Linux environments (SSH, Docker, Code Server) where iloom was previously unusable
• Thanks to @TickTockBent for the implementation and @rexsilex for the original architecture

Auto-Swarm: One Command from Plan to Execution

• il plan --auto-swarm chains the entire epic lifecycle — plan decomposition, epic workspace creation, and swarm launch — into a single unattended pipeline (#769)
• If the plan produces no child issues, it falls back to a normal autonomous loom
• VS Code extension can own the harness socket to manage the pipeline externally

PR Code Reviews with Inline Comments

• Agents can now post inline code review comments directly on GitHub pull requests, addressing specific lines of code rather than leaving generic top-level feedback (#819)
• Severity-based verdicts: Critical and Warning findings trigger REQUEST_CHANGES, suggestions are informational
• In one-shot mode, agents automatically triage and address existing review comments — fixing quick and moderate items, deferring complex or risky ones
• When opening a PR, agents check for unresolved review comments and work through them

Swarm Quality Presets in Configuration Wizard

• il init now offers swarm quality modes upfront: Maximum Quality (Opus everywhere), Balanced (Opus for analysis and workers, Sonnet for planning), and Fast & Cheap (Sonnet/Haiku) (#731, #777)
• Pick a preset and start swarming without manually configuring per-agent models

--complexity Flag to Skip Complexity Evaluation

• Tell iloom exactly how complex a task is instead of waiting for the evaluator: il start --complexity=simple or il spin --complexity=complex (#816)
• On il start, the override persists in loom metadata for future spins
• On il spin, it applies to the current session only

Configurable Per-Agent Swarm Models

• Set different AI models for each phase agent in swarm mode via per-agent swarmModel overrides, independent of non-swarm model settings (#736)

Post-Swarm Code Review Phase

• After all swarm child agents complete and merge, the orchestrator automatically runs a code review on the merged epic branch and auto-fixes high-confidence findings (#827)
• Configurable via spin.postSwarmReview setting (enabled by default)
• Non-blocking: review or fix failures don't prevent finalization
• Single-pass constraint prevents infinite review-fix loops

--create-only Flag for Workspace-Only Creation

• New convenience flag on il start that bundles --no-claude --no-code --no-terminal --no-dev-server — creates the isolated workspace without launching anything (#825)
• Useful for automation, scripting, and cases where you want to set up the environment before starting work

--json-stream for Headless Automation

• il rebase, il finish, and il commit now support --json-stream for JSONL streaming to stdout with progress on stderr — useful for CI pipelines and tooling integrations (#635)

il install-deps Command

• New utility command that detects and installs dependencies using iloom config files or lockfile detection, with support for pnpm, npm, and yarn (#780)

🐛 Bug Fixes

Swarm Reliability and Performance — Workers now produce descriptive commit messages instead of generic "fixes #NNN"; dependencies are installed after merging worker branches so subsequent workers have up-to-date packages; re-spinning skips done child issues so partial failures don't redo completed work; reduced per-agent overhead keeps worker context windows lean; il spin in child worktrees now shows a clear error instead of conflicting with the orchestrator; increased agent turn limits to prevent premature cutoffs on complex tasks; skipped redundant file validation in swarm mode to save 5-10 turns per implementation run (#781, #831, #830, #789, #792, #774, #860)

Agent Analysis Quality — Anti-confabulation rules prevent agents from fabricating cross-component values; anti-parroting rules ensure assumptions add genuine insight rather than restating the issue description; recap entries filter out agent infrastructure noise (#823, #828, #817)

Jira Wiki Markup Sanitization — Prevents Jira Wiki format from leaking into issue comments when Jira is configured as the issue tracker (#728)

Code Reviewer Works on All Branch Types — Previously only worked with uncommitted changes. Now diffs against the appropriate merge base for feature branches (#785)

Error Visibility When Claude Launch Fails — Error output from Claude is now properly surfaced instead of being silently swallowed (#786)

---

Full Changelog: v0.10.0...v0.11.1

v0.11.0

Release Notes (v0.10.0 → v0.11.0)

🎉 New Features

Linux, WSL, and tmux Terminal Support

• iloom now works on Linux with GUI terminals (gnome-terminal, konsole, xterm), WSL via Windows Terminal, and headless environments via tmux fallback (#796)
• Auto-detection: tries GUI terminals first, falls back to tmux when no display is available
• Unblocks il start on headless Linux environments (SSH, Docker, Code Server) where iloom was previously unusable
• Thanks to @TickTockBent for the implementation and @rexsilex for the original architecture

Auto-Swarm: One Command from Plan to Execution

• il plan --auto-swarm chains the entire epic lifecycle — plan decomposition, epic workspace creation, and swarm launch — into a single unattended pipeline (#769)
• If the plan produces no child issues, it falls back to a normal autonomous loom
• VS Code extension can own the harness socket to manage the pipeline externally

PR Code Reviews with Inline Comments

• Agents can now post inline code review comments directly on GitHub pull requests, addressing specific lines of code rather than leaving generic top-level feedback (#819)
• Severity-based verdicts: Critical and Warning findings trigger REQUEST_CHANGES, suggestions are informational
• In one-shot mode, agents automatically triage and address existing review comments — fixing quick and moderate items, deferring complex or risky ones
• When opening a PR, agents check for unresolved review comments and work through them

Swarm Quality Presets in Configuration Wizard

• il init now offers swarm quality modes upfront: Maximum Quality (Opus everywhere), Balanced (Opus for analysis and workers, Sonnet for planning), and Fast & Cheap (Sonnet/Haiku) (#731, #777)
• Pick a preset and start swarming without manually configuring per-agent models

--complexity Flag to Skip Complexity Evaluation

• Tell iloom exactly how complex a task is instead of waiting for the evaluator: il start --complexity=simple or il spin --complexity=complex (#816)
• On il start, the override persists in loom metadata for future spins
• On il spin, it applies to the current session only

Configurable Per-Agent Swarm Models

• Set different AI models for each phase agent in swarm mode via per-agent swarmModel overrides, independent of non-swarm model settings (#736)

Post-Swarm Code Review Phase

• After all swarm child agents complete and merge, the orchestrator automatically runs a code review on the merged epic branch and auto-fixes high-confidence findings (#827)
• Configurable via spin.postSwarmReview setting (enabled by default)
• Non-blocking: review or fix failures don't prevent finalization
• Single-pass constraint prevents infinite review-fix loops

--create-only Flag for Workspace-Only Creation

• New convenience flag on il start that bundles --no-claude --no-code --no-terminal --no-dev-server — creates the isolated workspace without launching anything (#825)
• Useful for automation, scripting, and cases where you want to set up the environment before starting work

--json-stream for Headless Automation

• il rebase, il finish, and il commit now support --json-stream for JSONL streaming to stdout with progress on stderr — useful for CI pipelines and tooling integrations (#635)

il install-deps Command

• New utility command that detects and installs dependencies using iloom config files or lockfile detection, with support for pnpm, npm, and yarn (#780)

🐛 Bug Fixes

Swarm Reliability and Performance — Workers now produce descriptive commit messages instead of generic "fixes #NNN"; dependencies are installed after merging worker branches so subsequent workers have up-to-date packages; re-spinning skips done child issues so partial failures don't redo completed work; reduced per-agent overhead keeps worker context windows lean; il spin in child worktrees now shows a clear error instead of conflicting with the orchestrator (#781, #831, #830, #789, #792, #774)

Agent Analysis Quality — Anti-confabulation rules prevent agents from fabricating cross-component values; anti-parroting rules ensure assumptions add genuine insight rather than restating the issue description; recap entries filter out agent infrastructure noise (#823, #828, #817)

Jira Wiki Markup Sanitization — Prevents Jira Wiki format from leaking into issue comments when Jira is configured as the issue tracker (#728)

Code Reviewer Works on All Branch Types — Previously only worked with uncommitted changes. Now diffs against the appropriate merge base for feature branches (#785)

Error Visibility When Claude Launch Fails — Error output from Claude is now properly surfaced instead of being silently swallowed (#786)

---

Full Changelog: v0.10.0...v0.11.0

v0.10.2

Release Notes (v0.9.0 → v0.10.2)

🎉 New Features

Swarm Mode: Parallel Agent Teams for Epics

The headline feature of this release. After decomposing an epic into child issues with il plan, running il start on the epic launches a team of AI agents that execute all children in parallel — each in its own isolated worktree, respecting dependency order. The VS Code extension visualizes swarm progress with a live Kanban board and dependency graph.

• Parallel execution with dependency awareness: Unblocked issues run simultaneously; newly unblocked issues launch as their dependencies finish
• Planning optimized for parallelism: The planner now favors contract-based parallelism — issues that share interfaces define the contract upfront and run simultaneously rather than sequentially (#654, #658)
• Rebase and merge: Completed child work is rebased and fast-forward merged into the epic branch for clean, linear history
• Incremental visibility in draft PR mode: The epic's draft PR updates after each child merge so you can track progress in real time (#681, #694)
• Failure isolation: Failed children don't block unrelated issues — they're preserved for debugging
• Automatic child cleanup: After each child merges, its worktree is archived and cleaned up. Use --skip-cleanup on il spin to keep them around (#661)
• Configurable models per phase: Set different AI models for swarm workers vs. normal looms, and even per phase agent (enhancer, analyzer, planner, implementer) within swarms (#653, #689, #735)
• Auto-detection: il start 100 detects child issues and prompts for epic mode. Use --epic to skip the prompt, --no-epic to force a normal loom (#619)
• Dependency installation after merge: When a worker branch is merged into the epic branch, dependencies are automatically installed so subsequent workers have up-to-date packages (#781)

Auto-Swarm: One Command from Plan to Execution

il plan --auto-swarm chains the entire epic lifecycle — plan decomposition, epic workspace creation, and swarm launch — into a single unattended pipeline. The planner signals completion via IPC, then iloom automatically runs il start --epic and il spin without human intervention. If the plan produces no child issues, it falls back to a normal autonomous loom (#769).

Swarm Quality Presets in the Configuration Wizard

il init now offers swarm quality modes upfront: Maximum Quality (Opus everywhere), Balanced (Opus for analysis, Sonnet for implementation), and Fast & Cheap (Sonnet/Haiku). Pick a preset and start swarming without manually configuring per-agent models (#731, #777).

--json-stream Flag for Headless Automation

il rebase, il finish, and il commit now support --json-stream for non-interactive use. Claude runs headless with JSONL streaming to stdout and progress redirected to stderr — useful for CI pipelines and tooling integrations (#635).

Jira Cloud Integration

iloom now supports Jira Cloud as a first-class issue tracker alongside GitHub and Linear (#588, #637).

• Full read/write on Jira issues with automatic Markdown-to-ADF conversion
• Child issue creation, dependency management, and workflow transitions
• il issues --sprint current and il issues --mine filtering for Jira boards
• Configurable issue types, done statuses, and transition mappings
• Automatic Jira Wiki markup sanitization prevents wiki format from leaking into comments when Jira is configured (#728)
• Thanks to @NoahCardoza for the foundational contribution

il issues --mine Works Everywhere

The --mine flag for filtering issues assigned to you now works across all providers — GitHub, Linear, and Jira (#606).

PR Review Comments Are Now Accessible to Agents

Agents can now read inline PR review comments (code review comments on specific files and lines), making it easy to address reviewer feedback directly (#715).

il finish Opens Your PR in the Browser

After marking a draft PR ready or creating a new PR, il finish now opens it in your browser so you can review it immediately. Disable with --no-browser or the mergeBehavior.openBrowserOnFinish setting (#673).

il finish Commits Without Prompting

The commit message review prompt is now skipped by default on il finish, reducing friction. Use --review when you want to inspect the message before committing (#686).

Loom Environment Variables

il shell and il dev-server now export ILOOM_LOOM (the loom identifier) and ILOOM_COLOR_HEX (the loom's assigned color) as environment variables. Use them to customize your shell prompt or tint your dev server UI (#587).

il install-deps Command

New utility command that detects and installs dependencies using iloom config files or lockfile detection. Supports pnpm, npm, and yarn with frozen lockfile semantics (#780).

Faster Utility Claude Calls

Branch name generation and commit message generation now run Claude in minimal mode, skipping MCP servers, hooks, and CLAUDE.md loading for noticeably faster startup on these lightweight tasks (#757).

Anonymous Usage Telemetry

Transparent, opt-out product analytics to help us understand which features are used and where users hit problems (#674).

• No personal information, repo names, code, or branch names are ever collected
• First-run disclosure with easy opt-out: il telemetry off
• il telemetry status to check current state

VS Code Extension Integration

• VS Code-aware wrap-up instructions: When running inside VS Code, agents now direct you to the Finish button in the iloom Explorer panel instead of CLI commands (#720)
• Zero-setup experience: il init --accept-defaults enables the VS Code extension to configure projects with a single click (#711)

🐛 Bug Fixes

Swarm worker agents following the workflow reliably — Multiple improvements to ensure worker agents delegate to phase sub-agents rather than attempting implementation directly. Workers now see their orchestrator identity before any other content and receive stripped-down task prompts that prevent "going rogue" (#752, #755, #772)

Planning produces better-structured epics — Phase 3 now enforces DAG-first, contract-first planning so agents design the dependency graph before writing issue details. Child issue bodies contain interface contracts only, not implementation instructions (#737, #775)

il spin blocked in child worktrees — Running il spin inside a child worktree of an epic now shows a clear error, since the swarm orchestrator manages those agents (#774)

Improved error messages when Claude launch fails — Error output from Claude is now properly surfaced instead of being silently swallowed (#786)

Parallel child worktree creation — Swarm setup now creates all child worktrees concurrently instead of one at a time, significantly speeding up epic initialization (#758)

Enhancer prompt corruption — Raw Handlebars template syntax was leaking into agent prompts when the enhancer loaded unprocessed templates (#607)

Stale rebase state — If a previous rebase was interrupted, starting a new one would fail. iloom now detects and aborts stale rebase state automatically (#663)

il finish installs dependencies after rebase — When main has new dependencies, validation no longer fails with missing module errors (#693)

Fork PR support — il start for PRs from forks no longer fails when the contributor's branch doesn't exist on origin (#594)

Child loom draft PRs target parent branch — Draft PRs for child looms now correctly target the parent loom's branch instead of main (#580)

Branch cleanup for draft PR looms — il cleanup no longer fails on draft PR branches that Git considers "not fully merged" (#576)

Cross-project loom collisions in il list — Issue numbers are now scoped by project path, preventing wrong metadata matches when multiple projects share issue numbers (#721)

il recap works for archived looms — You can now view recaps for looms that have been finished or cleaned up, not just active ones (#725)

---

Full Changelog: v0.9.0...v0.10.2