gald3r v2.0.1

Next Release — gald3r

Status: Draft — version number assigned at publish time.

Use this file for promotional, enticing copy that will headline the next GitHub release.
Technical detail belongs in CHANGELOG.md. User-facing how-to belongs in the
README.md pending section.

---

Headline

(One punchy line — what makes this release worth upgrading?)

---

What's coming

(Bullets that sell the release: outcomes, not implementation.)

---

Why upgrade now

(Optional — urgency, compatibility, or "if you skipped 2.0…")

---

See CHANGELOG.md for the full technical list.

gald3r v2.0.0

The gald3r engine release. gald3r gains a bundled, file-first Python core that backs every
system deterministically — while staying 100% markdown-on-disk. Existing installs keep working;
the engine is additive, and every slimmed component ships a no-engine fallback.

Added

• Bundled gald3r engine (.gald3r_sys/engine/) — a pure, file-first state backend for every
  system: tasks, bugs, features, goals, prds, ideas, vocab, constraints, subsystems, vault,
  release, workspace, and inbox. Mode-A: deterministic, no LLM, no network, no Docker. One
  prerequisite — uv.
• gald3r CLI (and python -m gald3r) — drive every system from the shell: gald3r task new,
  gald3r bug new, gald3r goal add, gald3r vault ingest, gald3r release new,
  gald3r workspace …, gald3r prompt get ….
• MCP server (gald3r mcp) — ~20 Model Context Protocol tools exposing the same operations to
  any MCP-capable agent.
• gald3r doctor — read-only health check (structure, per-system index integrity, skill
  frontmatter, .ps1 encoding) with an overall functionality score and a --fail-below CI gate.
• Engine-absorbed operations — five maintenance scripts reimplemented as pure engine verbs,
  each keeping its original .ps1 as a no-engine (L0) fallback: gald3r inbox · gald3r doctor ·
  gald3r platform status · gald3r tier show|set · gald3r sync --check|--apply (alias
  gald3r parity).
• Judgment / prompt layer — 15 reasoning assets (Norse persona, role briefs, review rubrics,
  marketing voice) served by the engine (gald3r prompt get role.code_reviewer), so a brief is
  authored once and shared across platforms.

Changed

• Thinned component shims — judgment skills and agents are slimmed to load their brief from the
  engine's prompt assets. Skills keep a full SKILL.full.md fallback; agents reference the shipped
  asset directly (no .full.md sidecar — it would register as a duplicate component).
• Task status vocabulary — task_file.v1.schema.yaml realigned to mirror the engine's enforced
  vocabulary (pending → in-progress → awaiting-verification → completed …). The YAML previously
  listed a never-implemented pipeline as "current" and the real vocabulary as "legacy."

Fixed

• Windows PowerShell 5.1 parse crash — 1,055 shipped .ps1 files were UTF-8 without a BOM, so
  powershell.exe mis-read multi-byte characters and failed to parse (including the installer
  itself). All BOM-protected (installer ASCII-cleaned); the build generators now emit safe .ps1
  and gald3r doctor flags any regression.
• Duplicate component names — removed the per-agent *.full.md sidecars and the deprecated
  g-skl-medkit (named g-skl-medic, colliding with the real skill). 106 skills + 13 agents now
  audit clean (no duplicate name:, no dangling shim references).
• doctor / bug sync index mis-parse — the id-scan matched the ## Next Bug ID: counter line
  (and any title mentioning it), producing false phantom/orphan rows and a non-converging
  bug sync. Anchored to the counter heading.
• Malformed component frontmatter — added missing name/description to 5 skills and agents.

Engineering

• 97 engine unit tests (pytest). The engine is the new source of truth; .gald3r_sys/schemas/
  mirrors it.

---

gald3r v1.10

release: gald3r v1.10 - Cursor + Claude unity edition (other platform…

gald3r v1.9.0

[1.9.0] - 2026-05-31 (Platforms/ Folder + Post-Push Verify + Plugin System Foundation)

Added

• Plugin system foundation (T1557–T1559): @g-plugin-install lands — install a gald3r plugin from a local path (GitHub-URL path implemented, not yet network-tested). Backed by ADR-015 (.gald3r_sys/docs/adr/ADR-015-plugin-system.md), a gald3r-plugin.yaml manifest schema + validator (.gald3r_sys/plugins/schema/), and a security-first installer: it validates the manifest, enforces gald3r_min_version, refuses to overwrite gald3r-core components (conflict-abort, ADR-015 D6), stamps installed components with plugin_source:, records a .gald3r_sys/plugins/installed.yaml ledger, and never auto-runs plugin lifecycle scripts (opt-in -RunInstallScript, previews first, ADR-015 D7). Remaining plugin commands (list/new/remove/update/registry) are in progress (T1560–T1568).
• scan_platform_docs.ps1 -WithHtml / -HappyPath (T1545): after a platform's PLATFORM_SPEC.md is updated by a doc scan, automatically regenerates its docs/platforms/<name>_guide.html via generate_platform_guide.ps1 (per-platform, in a child process so one guide's failure is non-fatal) and stamps last_html_generated: in the canonical spec for staleness visibility. Default (no -WithHtml) behavior is unchanged — no HTML regeneration. Closes the "happy path" gap where the spec and its HTML guide drifted out of sync.
• check_platform_status.ps1 -GenerateMatrix (T1543): auto-populates .gald3r/PLATFORM_CAPABILITY_MATRIX.md from the 23 canonical PLATFORM_SPEC.md ## Capability Summary tables (Hooks / Rules / Skills / Commands / MCP, plus a Docs-Fresh cell computed from each spec's crawl_max_age_days). Replaces the all-❓ placeholder matrix with real values, cross-checks every cell against the hand-verified PLATFORM_STATUS.md and warns on disagreement (never overwrites STATUS). Wired into platform_parity_sync.ps1 -ValidatePlatformSpecs so the matrix refreshes after each parity sync. First run: Updated 138 cells (45 ✅, 54 ⚠️, 33 ❌, 6 ❓).

Changed

• gald3r public repo restructured: 34 platform directories moved from repo root into platforms/ subfolder (gald3r/platforms/<name>/) for a cleaner root (T1556). platform_parity_sync.ps1 -SyncToGald3r updated to target platforms/ automatically. DISTRIBUTION_PLAYBOOK.md Step 5 updated to reflect new target path.
• platform_parity_sync.ps1 -ValidatePlatformSpecs / GeneratePlatformCats now discover platforms by scanning gald3r_template/.gald3r_sys/platforms/ dynamically rather than a hardcoded list (T1556 alignment).

Added

• action_scripts/post_push_verify.ps1 (T1572): 11-check post-push gate that verifies a release landed correctly — VERSION file, CHANGELOG entry + content depth, releases/ file, git remote tag, GitHub release existence + release-notes body populated, public gald3r repo VERSION, public gald3r README version mention, wiki file freshness, and a 10-pattern secrets scan of the release diff (AWS keys, GH tokens, private keys, password/API-key literals, bearer tokens, credential URLs, Slack tokens, DB connection strings). Code-file hits are FAIL; doc-file matches WARN with example suppression. Wired into DISTRIBUTION_PLAYBOOK.md Step 8b.

Fixed

Removed

• @g-kamikaze and @g-juggernaut commands (T1548): both were pure aliases for @g-mission with no added behavior — three names for one command added cognitive load without value. Deleted the command files (and their .github/prompts Copilot copies) across all platform targets, removed the alias references from g-mission and docs/wiki/Commands.md, and updated the parity-sync mission-command manifest. Use @g-mission directly.

---

gald3r v1.8.0

[1.8.0] - 2026-05-30 (Wiki Launch + GitHub Discussions + Test Harness + 35-Platform Sweep)

Added

• GitHub wiki launched at github.com/wrm3/gald3r/wiki (T1550): 7 pages auto-generated from .gald3r_sys/ — Home, Quickstart, Commands (179 @g-* commands indexed by category), Skills, Agents, Rules, Hooks. Sourced from canonical .gald3r_sys/ and staged in docs/wiki/. Full argument/syntax expansion tracked in T1551.
• GitHub Discussions enabled as the community Q&A and announcement channel.
• Batch-11 platform specs consolidated into the canonical gald3r_template/.gald3r_sys/platforms/.<platform>/PLATFORM_SPEC.md single-source-of-truth location (T1544): .deepcode, .kilo, .hermes, .codebuddy, .astrbot, .amp, .void, .continue, .kimi, .trae, .qoder, plus a .vibe stub spec that cross-links to .mistral. The canonical (clean UTF-8) .cursor/skills body was used; the divergent corrupted .claude/skills copies (BOM + mojibake) for 5 of them were discarded. The redundant PLATFORM_SPEC.md copies were removed from all 11 .claude/ and .cursor/ skill dirs (their SKILL.md is unchanged). platforms/ now holds 35 platform specs (23 original + 11 batch + .vibe).
• gald3r systems functional test harness custom_scripts/gald3r_system_test.ps1 (T1540): per-system PASS/PARTIAL/FAIL + overall "N% functional" score across 13 gald3r systems (Task, Bug, PLATFORM_SPEC, Parity, Hook Wiring, Git Hooks, Schema, Constraints, Subsystems, Skills, WPAC, Release, Encoding). Writes .gald3r/reports/system_test_*.md; -FailBelow <N> CI gate; -Json/-NoReport/-Systems flags. Added as the FUNCTIONAL (L0) operation in g-skl-test. Baselines in docs/TEST_BASELINE.md; CI guidance in docs/20260530_171707_Cursor_SYSTEM_TEST_CI_INTEGRATION.md.
• Two-layer release system: action_scripts/pre_release_audit.ps1, bump_identity_versions.ps1, release_gald3r_public.ps1, release_config.json (T1528-T1530)
• Native tier graduation scripts in g-skl-release/scripts/: graduate_to_public.ps1 (scrub + carry modes), graduate_to_test.ps1, scrub_public_tree.ps1, read_tier_config.ps1
• Restored custom_scripts/export_slim_template_repo.ps1; @g-release-sync command
• Framework constraints C-041..C-043 (destructive git gate, content scrub, public main-only)

Changed

• custom_scripts/scan_platform_docs.ps1 now discovers platforms by scanning gald3r_template/.gald3r_sys/platforms/ for any dir containing a PLATFORM_SPEC.md (no hardcoded 23-platform list; T1544 AC6), reads the canonical PLATFORM_SPEC.md as the capability record, and updates last_doc_scan in that spec plus any surviving SKILL.md copies. Dir-to-name normalization handles .replit-gald3r->replit and .kilo->kilo-code. Discovery verified at 35 platforms under powershell.exe 5.1.
• DISTRIBUTION_PLAYBOOK.md, BRANCH_TIER_POLICY.md, GIT_TIER_MODEL.md -- guarded public release path; test_to_public_history: scrub|carry in AGENT_CONFIG
• Branch model: feature-branches-only -> main (USER-SAFETY, T1535). Retired the long-lived dev/test promotion branches that were the root cause of repeated history-loss incidents (BUG-099 class: stale dev auto-merge target). g-rl-02-git_workflow rewritten to feature-branches-only; g-go/g-go-go auto-merge default flipped dev -> main; gald3r_worktree.ps1 -TargetBranch default dev -> main; development.yaml default_branch dev -> main; g-skl-setup now checks for main instead of creating a dev branch. There is no longer any long-lived dev/test branch in the gald3r workflow.

Fixed

• Eliminated the stale-dev auto-merge data-loss hazard (BUG-099, USER-SAFETY): gald3r no longer defaults swarm/autopilot auto-merge to a long-lived dev integration branch. Existing user repos still on the old model are offered a safe, confirmation-gated migration (promote dev -> main, then retire dev) — never an automatic branch delete.

---

gald3r v1.7.0

Released: 2026-05-28
Version: 1.7.0
Type: Minor -- new features, no breaking changes

---

Highlights

34 AI Coding Tools -- Full Sweep

11 new platforms added in this release: Kimi Code (Moonshot AI), TRAE (ByteDance),
Qoder (Alibaba), Amp Code (Sourcegraph), Continue, Void Editor, Deep Code (DeepSeek),
Kilo Code, Hermes (Nous Research), CodeBuddy (Tencent), AstrBot. All 34 platforms
verified QA-passing.

g-go-go Coding Swarm Fix

g-go-go Phase 1 now explicitly invokes g-go-code-swarm (N parallel coders) instead
of running one coder at a time. Phase 2 was already parallel (review swarm) -- Phase 1
was silently sequential. This was the root cause of the framework feeling slow over the
past week. Added --no-code-swarm opt-out flag.

Context-Aware Throttle ON by Default

--context-aware is now the default. Proactively reduces N under context pressure
instead of triggering a CONTEXT WINDOW PANIC stop. --no-context-aware to opt out.

Platform Restructure (T1522)

All 34 platform templates now live at repo root (cursor/, claude/, etc.) -- no more
platforms/<name>/ nesting. Each folder is ready to drop directly into your project.

---

Full Change List

See CHANGELOG.md ## [1.7.0] section for the complete detailed list.

---

Install / Upgrade

Pick your platform folder and copy its contents to your project root:

# Example for Cursor
cp -r cursor/ /your/project/

For a fresh install see instructions_new_project.md.
For upgrading an existing install see instructions_existing_project.md.

---

Platform Count

Tier	Count	Platforms
Tier 1 (Fully Supported)	11	Cursor, Claude Code, Copilot, Windsurf, Cline, Roo, Codex, CodeBuddy, Amp, Continue, Kimi
Tier 2 (Community)	11	Aider, Augment, Goose, Warp, OpenHands, Kiro, Kiro CLI, Junie, Replit, Gemini, Kilo Code, Qoder
Tier 3 (Experimental)	9	Mistral, Antigravity, OpenClaw, Qwen, SubQ, DeepCode, Hermes, AstrBot, Void, TRAE
Total	34

gald3r v1.6.0 -- WPAC v1.6, Schema System, Ship Skill, Encoding Normalization

Released: 2026-05-25
Version: 1.6.0
Type: Minor -- new features, no breaking changes
Previous release: v1.5.0

---

Highlights

Schema enforcement system

Every .gald3r/ file now carries version provenance, validated against 15 versioned
schema definitions. A read-only session-start probe flags drift in under a second, and a
data-preserving migration engine (migrate_schemas.ps1) upgrades older files in place.
g-medic gained an L1 validation layer that auto-fixes common schema gaps.

Semantic versioning and release management

New g-skl-ship skill and @g-ship command promote your CHANGELOG [Unreleased] section
to a real version: it bumps VERSION, tags the release, and can publish. Task and bug
completion now feed CHANGELOG entries automatically.

Dedicated bug-fix pipeline

@g-go-bugs (and its swarm variant) is an autopilot built only for bugs: it reproduces,
fixes, writes a regression test, and runs an adversarial cold review -- working through
your backlog in severity order, critical first.

Encoding normalization

A new pre-commit and stop-event hook normalizes line endings and BOM policy, with a
content-aware guard that leaves real binary files untouched. Shipped with a .gitattributes
scaffold and a one-command git-hooks installer so fresh installs are protected by default.

Cross-project promotion lifecycle

Workspace members can now graduate from a lightweight "controlled member" to a fully
independent "autonomous child" via @g-wpac-promote, with a formal, dry-run-first
lifecycle instead of hand-editing files.

Autopilot robustness

g-go-go gained a context-aware mode that shrinks its parallel batch size under context
pressure instead of stopping mid-run, plus a stop-detection guard that resumes a stalled
loop when it halts without an authorized reason.

---

Notable Fixes

• Cross-platform file corruption from an early coordination transition was cleaned up
  across all framework trees, and a reusable cleanup script was added.
• Several PowerShell hook parse and compatibility issues resolved.
• Root VERSION file and .gald3r/releases/ history are now created and backfilled by
  both setup and @g-update.
• Context Budget and Conflict Pattern hard rules restored to the core enforcement ruleset.

---

Install / Upgrade

From v1.5.x, run in your project:

# Recommended
@g-update --apply

# Or migrate schemas directly (idempotent, safe to re-run)
.gald3r_sys\scripts\migrate_schemas.ps1 -Apply

For a fresh install see instructions_new_project.md.

gald3r v1.5.2 -- HTML/JSON/TOON output, memory compression, vocab, auto-skill

What's New in v1.5.2

Added

• HTML output + theme system (g-skl-html-output, --html flag): human-facing reports (g-status, g-review, g-qa-report) now render as self-contained themed HTML. Three built-in themes: gald3r-dark (Catppuccin Mocha, default), gald3r-light (Latte), gald3r-mocha (sepia). Opt-in via output_format: html in AGENT_CONFIG. Coordination files stay markdown.
• JSON output mode (g-skl-json-output, --json flag): structured {gald3r_version, generated_at, command, schema, data} envelope for scripting, CI gates, and dashboards.
• TOON output mode (g-skl-toon-output, --toon flag): Token-Oriented Object Notation — compact, lossless, LLM-friendly tabular arrays (~40% smaller than JSON).
• Theme editor (g-skl-theme-editor, @g-theme-edit): file-first theme editor for creating and editing gald3r HTML themes. Supports import/export of :root token blocks.
• Memory compression (g-skl-compress-memory, @g-compress-memory): compact session memory and learned-facts entries to reduce context bloat while preserving key insights.
• Vocabulary management (@g-vocab-add, @g-vocab-list, @g-vocab-search): define project-specific abbreviations that agents read at session start.
• Skill review command (@g-skill-review): list, review, and selectively promote auto-proposed skill drafts. Human approval required before promotion.
• Auto-skill-generation (--propose-skill on @g-go-review): optionally drafts a SKILL.md after a PASS verdict when a novel, generalizable pattern is found. Off by default.
• @g-mission improvements: resume --budget N flag, drain queue scan order, 75% context checkpoint threshold, mandatory split for oversized tasks, cross-repo task fix in --until-empty, autonomous push gate, and session-end framing corrections.

---

Full changelog: CHANGELOG.md

gald3r v1.5.1

Added

• docs/PLUGINS.md: plugin author guide covering skill pack structure, distribution, and naming guidelines for third-party extensions.
• README "Plugins & Skill Packs" section with naming guideline (plugins are separate works).
• NOTICE clarified with plugin naming guideline.

gald3r v1.5.0 -- Platform Framework Architecture (.gald3r_sys/)

Released: 2026-05-10
Version: 1.5.0 (Maestro Harvest)
Type: Minor -- new features, no breaking changes
Previous release: v1.4.0

---

Highlights

New security and quality skills

• g-skl-security-scan -- a two-phase SAST scanner. A fast, free regex pass surfaces
  candidates across 12 vulnerability categories (hardcoded credentials, SQL injection,
  eval/exec, weak crypto, path traversal, CORS wildcards, and more), then a focused LLM
  pass analyzes only the flagged batches. Complements g-skl-code-review.
• g-skl-context-builder -- assembles a token-budgeted context block on the fly from
  live .gald3r/ state (active tasks, constraints, relevant subsystems, recent memory),
  so agents start work with the right context and nothing more.
• g-skl-delegate -- an engineering-team delegation workflow: task-brief templates,
  code-review request templates, quality-gate checklists, and clean handoff protocols.

Recon and research suite

A unified research family replaces the older one-off ingest skills: capture a GitHub
repo, a URL, a docs site, a YouTube transcript, or a local file into your vault, then run
deep analysis and apply the findings as gald3r artifacts. Includes a similarity-risk
field so harvested patterns carry IP provenance from capture through apply.

More coding-agent runtimes

• g-skl-cli-jcode -- documents the jcode Rust agent as a low-overhead local runtime
  (millisecond startup, local embeddings via Ollama) alongside the existing CLI packs.
• g-skl-comfyui -- local GPU image and video generation via ComfyUI (SDXL,
  AnimateDiff) with zero cloud cost.

Pipeline and workflow improvements

• g-go-code context-prep can now query a code graph (g-skl-graphify) for far cheaper
  architecture lookups, with graceful fallback to grep-based prep.
• g-go --swarm gains file-lock manifests so parallel agents never edit the same files.
• Mid-flight @g-steer and @g-queue for in-progress worktree sessions, plus
  @g-go-code --resume crash recovery via checkpoint artifacts.
• New @g-triage command turns unstructured input (emails, Slack, meeting notes) into
  routed gald3r items behind a hard human-approval gate.
• AGENT_CONFIG.md harness tuning guide (context budgets, temperature presets, retries).

---

Install / Upgrade

Pick your platform folder and copy its contents to your project root:

# Example for Cursor
cp -r cursor/ /your/project/

For a fresh install see instructions_new_project.md.
For upgrading an existing install see instructions_existing_project.md.