wgm v0.4

What's Changed

• docs(references): add local-models.md token-budget playbook for small-context models by @SchwartzKamel in #64
• feat(evals): SkillOpt-inspired live grading + non-regression gate for evals.json by @SchwartzKamel in #65
• docs(audit): consolidate PRs #62-#65 findings + land agent-action fixes by @SchwartzKamel in #66
• feat: harden loop and audit evidence by @SchwartzKamel in #97
• docs: record WGM growth stage by @SchwartzKamel in #98
• feat: operationalize WGM growth opportunities by @SchwartzKamel in #99
• docs: record growth delivery state by @SchwartzKamel in #100

Full Changelog: v0.3...v0.4

wgm v0.3

What's Changed

• docs: operator executive overviews + competitive roadmap by @SchwartzKamel in #1
• feat: Tier 1 governance & discipline (constitution · analyze gate · no-placeholder · guardrails) by @SchwartzKamel in #2
• feat: cross-iteration memories (.wgm/memories.md) — roadmap Tier 2 #6 by @SchwartzKamel in #4
• feat(loop): hard operational limits + lifecycle hooks (roadmap Tier 2 #7) by @SchwartzKamel in #3
• chore(ci+docs): add actionlint workflow; sync roadmap & doc statuses by @SchwartzKamel in #5
• ci: add release workflow for tagged releases by @SchwartzKamel in #6
• docs: make the bundled Ralph loop usable from any project by @SchwartzKamel in #7
• feat: scale-adaptive tracks (roadmap Tier 2 #8) — built via the Ralph loop by @SchwartzKamel in #8
• feat: custom subagent archetypes + two-stage review (roadmap Tier 3 #9) by @SchwartzKamel in #9
• feat(loop): declarative wgm.yml project gates (roadmap Tier 2 #5) by @SchwartzKamel in #10
• docs: track wgm vs the Ralph ecosystem (awesome-ralph) by @SchwartzKamel in #11
• docs(landscape): add elves as nearest rival + track github.com/topics/ralph-loop by @SchwartzKamel in #12
• docs(references): token-economy note (gleaned from agent-smith's kanji mapping) by @SchwartzKamel in #13
• feat(token-economy): CONTEXT.md glossary + compact agent-only single-token encoding (Tier 3 #10) by @SchwartzKamel in #14
• feat(subagents): complete the 6-role swarm + optimize the roster by @SchwartzKamel in #15
• feat(self-improvement): the wgm growth flywheel — harvest, report, promote by @SchwartzKamel in #16
• feat(loop): context rotation + summarization at a token budget by @SchwartzKamel in #17
• feat(loop): retry with backoff/jitter + consecutive-failure circuit breaker by @SchwartzKamel in #18
• feat(loop): per-iteration metrics ledger + pluggable cost hook by @SchwartzKamel in #19
• feat(review): dissent preservation in the two-stage review by @SchwartzKamel in #20
• feat(spec): EARS notation for unambiguous, testable acceptance criteria by @SchwartzKamel in #21
• feat(swarm): parallel git-worktree streams (scripts/swarm.sh) by @SchwartzKamel in #22
• feat(dx): Makefile so make update / make validate just work by @SchwartzKamel in #23
• docs(landscape): track agent-skills + ai-coding adjacent ecosystems by @SchwartzKamel in #24
• docs: cover new features in README + indexes (with a junior-engineer clarity pass) by @SchwartzKamel in #25
• feat(dx): make update pulls latest; installers consume the release asset by @SchwartzKamel in #26
• docs(SKILL): compress protocol prose to cut skill-injection token cost by @SchwartzKamel in #27
• feat(plugin): add sofaking plugin integration scaffolding by @SchwartzKamel in #28
• feat(docs-audit): mandatory 4-persona docs audit + operator SOP by @SchwartzKamel in #38
• feat(heuristics): harvest self-improvement batch 1 (issues #29-#37) by @SchwartzKamel in #39
• feat(loop): add --max-cost cost ceiling to scripts/loop.sh by @SchwartzKamel in #40
• docs(audit): fresh post-merge docs-audit pass + fix a real gap found by @SchwartzKamel in #41
• docs(trigger-eval): add a skill-behaviour should/should-not fixture by @SchwartzKamel in #42
• feat(self-improvement): assimilate 3 GitHub agent-skill findings by @SchwartzKamel in #43
• docs: assimilate Spec Kit and GSD grill/plan heuristics by @SchwartzKamel in #45
• feat(references): assimilate BMAD review + eval automation findings by @SchwartzKamel in #44
• docs(audit): consolidate 4-persona review of #43, re-scoped to cover #41-#45 by @SchwartzKamel in #46
• docs: assimilate Kiro and aider grill/loop findings by @SchwartzKamel in #47
• Assimilate round 3 scoring, gene, and subagent findings by @SchwartzKamel in #48
• docs(audit): consolidate round-3 (PRs #47-#48) docs-audit findings by @SchwartzKamel in #49
• docs: extend Ralph loop guidance by @SchwartzKamel in #50
• Document optional memory patterns and sprint status template by @SchwartzKamel in #52
• docs: add operator tooling handoff patterns by @SchwartzKamel in #51
• docs: add ADR and governance-discovery references by @SchwartzKamel in #53
• docs(audit): consolidate round-4 (PRs #50-#53) docs-audit findings by @SchwartzKamel in #54
• Prefer Ralph-full by default + a disk-conscious local devcontainer sandbox by @SchwartzKamel in #55
• docs: honest growth health-check + self-improvement guardrail + 2 real dogfood harvests by @SchwartzKamel in #58
• feat(self-improvement): Hive Growth Loop — swarm learning, issue intake, consented auto-report by @SchwartzKamel in #61
• fix(hive-growth-loop): docs-audit swarm findings + real bug fixes from PR #61 by @SchwartzKamel in #62
• docs(heuristics): harvest [learn] issue #60 — PR CI green is not proof of deploy by @SchwartzKamel in #63

New Contributors

• @SchwartzKamel made their first contribution in #1

Full Changelog: https://github.com/agent-frontier/wgm/commits/v0.3